Liste der Fachpublikationen

Ergebnisse 1 - 157 von 157
Show publication details
Getto, Roman; Fina, Kenten; Jarms, Lennart; Kuijper, Arjan; Fellner, Dieter W.

3D Object Classification and Parameter Estimation based on Parametric Procedural Models

2018

WSCG 2018. Full Papers Proceedings

International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG) <26, 2018, Plzen, Czech Republic>

Computer Science Research Notes (CSRN)
2801

Classifying and gathering additional information about an unknown 3D objects is dependent on having a large amount of learning data. We propose to use procedural models as data foundation for this task. In our method we (semi-)automatically define parameters for a procedural model constructed with a modeling tool. Then we use the procedural models to classify an object and also automatically estimate the best parameters. We use a standard convolutional neural network and three different object similarity measures to estimate the best parameters at each degree of detail. We evaluate all steps of our approach using several procedural models and show that we can achieve high classification accuracy and meaningful parameters for unknown objects.

  • 978-80-86943-40-4
Show publication details

3D Printing Spatially Varying Color and Translucency

2018

ACM Transactions on Graphics

We present an efficient and scalable pipeline for fabricating full-colored objects with spatially-varying translucency from practical and accessible input data via multi-material 3D printing. Observing that the costs associated with BSSRDF measurement and processing are high, the range of 3D printable BSSRDFs are severely limited, and that the human visual system relies only on simple high-level cues to perceive translucency, we propose a method based on reproducing perceptual translucency cues. The input to our pipeline is an RGBA signal defined on the surface of an object, making our approach accessible and practical for designers. We propose a framework for extending standard color management and profiling to combined color and translucency management using a gamut correspondence strategy we call opaque relative processing. We present an efficient streaming method to compute voxel-level material arrangements, achieving both realistic reproduction of measured translucent materials and artistic effects involving multiple fully or partially transparent geometries.

Show publication details
Krispel, Ulrich; Fellner, Dieter W.; Ullrich, Torsten

A Benchmark for Distance Measurements

2018

2018 International Conference on Cyberworlds

International Conference on Cyberworlds (CW) <2018, Singapore>

The need to analyze and visualize distances between objects arises in many use cases. Although the problem to calculate the distance between two polygonal objects may sound simple, real-world scenarios with large models will always be challenging, but optimization techniques – such as space partitioning – can reduce the complexity of the average case significantly. Our contribution to this problem is a publicly available benchmark to compare distance calculation algorithms. Furthermore, we evaluated the two most important techniques (hierarchical tree structures versus grid-based approaches).

  • 978-1-5386-7315-7
Show publication details
Sarcher, Julian; Scheglmann, Christian; Zoellner, Alexander; Dolereit, Tim; Schäferling, Michael; Vahl, Matthias; Kiefer, Gundolf

A Configurable Framework for Hough-Transform-Based Embedded Object Recognition Systems

2018

2018 IEEE 29th International Conference on Application-specific Systems, Architectures and Processors (ASAP)

International Conference on Application-specific Systems, Architectures and Processors (ASAP) <29, 2018, Milan, Italy>

Real-time object recognition on low-power embedded devices is a widely requested task, needed in manifold applications. However, it is still a demanding challenge to achieve desired performance goals. For example, for advanced driver assistance systems (ADAS) or autonomously driven cars, object recognition and lane detection are indispensable tasks. Another field of application is the continuous retrieval of the construction progress on-site for validation of the construction site status, by detecting installed components using a given CAD model. This paper presents a framework for highly customizable object detection systems implemented on a single heterogeneous computing chip leveraging FPGA logic and standard processors. The FPGA logic is used to implement a custom variation of the Hough Transform and further image processing tasks efficiently. The dedicated logic is supplemented with a software stack, which consists of a Linux operating system, including hardware access drivers, as well as high-level libraries like OpenCV and Robot Operating System (ROS) - all running on the same device. The capabilities of the system are demonstrated for three application scenarios, namely race track recognition, lane recognition and object detection tasks performed within a construction assistance system.

  • 978-1-5386-7479-6
Show publication details
Ding, Peng; Zhang, Ye; Deng, Wei-Jian; Jia, Ping; Kuijper, Arjan

A Light and Faster Regional Convolutional Neural Network for Object Detection in Optical Remote Sensing Images

2018

ISPRS Journal of Photogrammetry and Remote Sensing

Detection of objects from satellite optical remote sensing images is very important for many commercial and governmental applications. With the development of deep convolutional neural networks (deep CNNs), the field of object detection has seen tremendous advances. Currently, objects in satellite remote sensing images can be detected using deep CNNs. In general, optical remote sensing images contain many dense and small objects, and the use of the original Faster Regional CNN framework does not yield a suitably high precision. Therefore, after careful analysis we adopt dense convoluted networks, a multi-scale representation and various combinations of improvement schemes to enhance the structure of the base VGG16-Net for improving the precision. We propose an approach to reduce the test-time (detection time) and memory requirements. To validate the effectiveness of our approach, we perform experiments using satellite remote sensing image datasets of aircraft and automobiles. The results show that the improved network structure can detect objects in satellite optical remote sensing images more accurately and efficiently.

Show publication details
Siegmund, Dirk; Dev, Sudeep; Fu, Biying; Scheller, Doreen; Braun, Andreas

A Look at Feet: Recognizing Tailgating via Capacitive Sensing

2018

Distributed, Ambient, and Pervasive Interactions: Technologies and Contexts

International Conference on Distributed, Ambient and Pervasive Interactions (DAPI) <6, 2018, Las Vegas, NV, USA>

At many every day places, the ability to be reliably able to determine how many individuals are within an automated access control area, is of great importance. Especially in high-security areas such as banks and at country borders, access systems like mantraps or drop-arm turnstiles serve this purpose. These automated systems are designed to ensure that only one person can pass through a particular transit area at a time. State of the art systems use camera systems mounted in the ceiling to detect people sneaking in behind authorized individuals to pass through the transit space (tailgating attacks). Our novel method is inspired by recently achieved results in capacitive in-door-localization. Instead of estimating the position of humans, the pervasive capacitance of feet in the transit space is measured to detect tailgating attacks. We explore suitable sensing techniques and sensor-grid layout to be used for that application. In contrast to existing work, we use machine learning techniques for classification of the sensor’s feature vector. The performance is evaluated on hardware-level, by defining its physical effectiveness. Tests with simulated attacks show its performance in comparison with competitive camera-image methods. Our method provides verification of tailgating attacks with an equal-error-rate of 3.5%, which outperforms other methods. We conclude with an evaluation of the amount of data needed for classification and highlight the usefulness of this method when combined with other imaging techniques.

Show publication details
Krämer, Michel; Fellner, Dieter W. [Referent]; Boehm, Jan [Referent]

A Microservice Architecture for the Processing of Large Geospatial Data in the Cloud

2018

Darmstadt, TU, Diss., 2018

With the growing number of devices that can collect spatiotemporal information, as well as the improving quality of sensors, the geospatial data volume increases constantly. Before the raw collected data can be used, it has to be processed. Currently, expert users are still relying on desktop-based Geographic Information Systems to perform processing workflows. However, the volume of geospatial data and the complexity of processing algorithms exceeds the capacities of their workstations. There is a paradigm shift from desktop solutions towards the Cloud, which offers virtually unlimited storage space and computational power, but developers of processing algorithms often have no background in computer science and hence no expertise in Cloud Computing. Our research hypothesis is that a microservice architecture and Domain-Specific Languages can be used to orchestrate existing geospatial processing algorithms, and to compose and execute geospatial workflows in a Cloud environment for efficient application development and enhanced stakeholder experience. We present a software architecture that contains extension points for processing algorithms (or microservices), a workflow management component for distributed service orchestration, and a workflow editor based on a Domain-Specific Language. The main aim is to provide both users and developers with the means to leverage the possibilities of the Cloud, without requiring them to have a deep knowledge of distributed computing. In order to conduct our research, we follow the Design Science Research Methodology. We perform an analysis of the problem domain and collect requirements as well as quality attributes for our architecture. To meet our research objectives, we design the architecture and develop approaches to workflow management and workflow modelling. We demonstrate the utility of our solution by applying it to two real-world use cases and evaluate the quality of our architecture based on defined scenarios. Finally, we critically discuss our results. Our contributions to the scientific community can be classified into three pillars. We present a scalable and modifiable microservice architecture for geospatial processing that supports distributed development and has a high availability. Further, we present novel approaches to service integration and orchestration in the Cloud as well as rule-based and dynamic workflow management without a priori design-time knowledge. For the workflow modelling we create a Domain-Specific Language that is based on a novel language design method.

Show publication details
Ma, Jingting; Lin, Feng; Wesarg, Stefan; Erdt, Marius

A Novel Bayesian Model Incorporating Deep Neural Network and Statistical Shape Model for Pancreas Segmentation

2018

Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: Part IV

International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) <21, 2018, Granada, Spain>

Lecture Notes in Computer Science (LNCS)
11073

Deep neural networks have achieved significant success in medical image segmentation in recent years. However, poor contrast to surrounding tissues and high flexibility of anatomical structure of the interest object are still challenges. On the other hand, statistical shape model based approaches have demonstrated promising performance on exploiting complex shape variabilities but they are sensitive to localization and initialization. This motivates us to leverage the rich shape priors learned from statistical shape models to improve the segmentation of deep neural networks. In this work, we propose a novel Bayesian model incorporating the segmentation results from both deep neural network and statistical shape model for segmentation. In evaluation, experiments are performed on 82 CT datasets of the challenging public NIH pancreas dataset. We report 85.32 % of the mean DSC that outperforms the state-of-the-art and approximately 12 % improvement from the predicted segment of deep neural network.

  • 978-3-030-00936-6
Show publication details
Balreira, Dennis G.; Walter, Marcelo; Fellner, Dieter W.

A Survey of the Contents in Introductory Computer Graphics Courses

2018

Computers & Graphics

Computer Graphics is a very active field, with new knowledge being published every day at a high rate. There is, therefore, the pressure to regularly review our teaching contents and adjust accordingly. Among the courses on a standard curriculum, the introductory Computer Graphics course is very often the door for students into the exciting area of Computer Graphics. It is also the opportunity to attract and engage the best talent for the field. In this paper, we address the question of content in the introductory Com- puter Graphics course as a community. Our main motivation was to find out what our peers are teaching in this first course and use this knowledge to ease the redesign of our introductory course. We have surveyed 28 introductory Computer Graphics undergraduate courses from higher level educational insti- tutions from around the world. We have asked the instructors of these courses to send us data on their courses, such as the weekly list of topics, and others such as textbooks. We gathered and processed this data using a bottom-up approach. The final top-level list of subjects and percentages for the introductory Computer Graphics courses, following the knowledge units defined in the 2013 ACM/IEEE recommenda- tion, is as follows: Rendering (71.3%), Geometric Modeling (17.4%), Animation (7.8%), Fundamentals (3.0%), and Visualization (0.5%). We believe this survey will be helpful for institutions considering designing a new introductory course from scratch or redesigning an existing one, by providing the current state-of- practice of top Computer Graphics institutions around the world.

Show publication details

A Universal, Closed-form Approach for Absolute Pose Problems

2018

Computer Vision and Image Understanding

We propose a general approach for absolute pose problems including the well known perspective-n-point (PnP) problem, its generalized variant (GPnP) with and without scale, and the pose from 2D line correspondences (PnL). These have received a tremendous attention in the computer vision community during the last decades. However, it was only recently that efficient, globally optimal, closed-form solutions have been proposed, which can handle arbitrary numbers of correspondences including minimal configurations as well as over-constrained cases with linear complexity. We follow the general scheme by eliminating the linear parameters first, which results in a least squares error function that only depends on the non-linear rotation and a small symmetric coefficient matrix of fixed size. Then, in a second step the rotation is solved with algorithms which are derived using methods from algebraic geometry such as the Gröbner basis method. We propose a unified formulation based on a representation with orthogonal complements which allows to combine different types of constraints elegantly in one single framework. We show that with our unified formulation existing polynomial solvers can be interchangeably applied to problem instances other than those they were originally proposed for. It becomes possible to compare them on various registrations problems with respect to accuracy, numerical stability, and computational speed. Our compression procedure not only preserves linear complexity, it is even faster than previous formulations. For the second step we also derive an own algebraic equation solver, which can additionally handle the registration from 3D point-to-point correspondences, where other rotation solvers fail. Finally, we also present a marker-based SLAM approach with automatic registration to a target coordinate system based on partial and distributed reference information. It represents an application example that goes beyond classical camera pose estimation from image measurements and also serves for evaluation on real data.

Show publication details
Dolereit, Tim; Urban, Bodo [Gutachter]; Koch, Reinhard [Gutachter]; Lukas, Uwe von [Gutachter]

A Virtual Object Point Model for the Calibration of Underwater Stereo Cameras to Recover Accurate 3D Information

2018

Rostock, Univ., Diss., 2018

Computer vision aims at processing information that is recorded with the aid of imaging devices. Compared to conventional computer vision in air, underwater computer vision is at an early stage of development. Water is a challenging environment, which, apart from the attenuation of light by absorption and scattering, is characterized in particular by the refraction of light at media transitions. Refractive effects are a problem for the recovery of metric 3D structure from underwater image data. The focus of this thesis is on recovering accurate 3D information from underwater images. In this thesis, the concepts of stereo 3D reconstruction in air will be extended for underwater environments by an explicit consideration of refractive effects. These concepts comprise projections for coordinate transformations between 3D and 2D spaces, calibration of the imaging device and the actual recovery of 3D coordinates. A novel model, named Virtual Object Point model, was developed to realize the necessary extensions. It is the cornerstone of this thesis, is characterized by its integrability into the geometric sub-process of underwater image formation and enables novel developments based on this. Within underwater stereo 3D reconstruction, the focus of this thesis is on the refractive calibration of a special class of underwater imaging systems. These systems, named Shared Flat Refractive System (SFRS), consist of a stereo camera and a viewing window, which is a single flat, transparent, refractive interface. The geometric sub-process of underwater image formation is modeled by the parameters of the SFRS with an explicit consideration of refractive effects. The performed investigations have shown that the particularities of the SFRS are beneficial for refractive calibrations and worth to adjust the basic design of the imaging system accordingly. A major contribution of this thesis is the development of various approaches for the calibration of the parameters of a SFRS. These approaches are characterized by the fact that part of the refractive parameters are determined by linear optimization. Furthermore, another crucial goal of this thesis, to relax or even eliminate restrictions due to the cumbersome and time-consuming procedure of handling special calibration objects under water, was achieved. Thus, the developed approaches for the refractive calibration of a SFRS show different degrees of dependence on calibration objects culminating in a full independence from those.

Show publication details
Smith, Neil; Moehrle, Nils; Goesele, Michael; Heidrich, Wolfgang

Aerial Path Planning for Urban Scene Reconstruction: A Continuous Optimization Method and Benchmark

2018

ACM Transactions on Graphics

Small unmanned aerial vehicles (UAVs) are ideal capturing devices for high-resolution urban 3D reconstructions using multi-view stereo. Nevertheless, practical considerations such as safety usually mean that access to the scan target is often only available for a short amount of time, especially in urban environments. It therefore becomes crucial to perform both view and path planning to minimize flight time while ensuring complete and accurate reconstructions. In this work, we address the challenge of automatic view and path planning for UAV-based aerial imaging with the goal of urban reconstruction from multi-view stereo. To this end, we develop a novel continuous optimization approach using heuristics for multi-view stereo reconstruction quality and apply it to the problem of path planning. Even for large scan areas, our method generates paths in only a few minutes, and is therefore ideally suited for deployment in the field. To evaluate our method, we introduce and describe a detailed benchmark dataset for UAV path planning in urban environments which can also be used to evaluate future research efforts on this topic. Using this dataset and both synthetic and real data, we demonstrate survey-grade urban reconstructions with ground resolutions of 1 cm or better on large areas (30 000m2).

Show publication details

An Integrated Deep Neural Network for Defect Detection in Dynamic Textile Textures

2018

Progress in Artificial Intelligence and Pattern Recognition

International Workshop on Artificial Intelligence and Pattern Recognition (IWAIPR) <6, 2018, Havana, Cuba>

Lecture Notes in Computer Science (LNCS)
11047

This paper presents a comprehensive defect detection method for two common fabric defects groups. Most existing systems require textiles to be spread out in order to detect defects. This method can be applied when the textiles are not spread out and does not require any pre- processing. The deep learning architecture we present is based on transfer learning and localizes and recognizes cuts, holes and stain defects. Classification and localization is combined into a single system combining two different networks. The experiments this paper presents show that even without adding depth information, the network was able to distinguish between stain and shadow. This method has been successful even for textiles in voluminous shape and is less computationally intensive than other state-of-the-art methods.

  • 978-3-030-01131-4
Show publication details
Scheller, Doreen; Bauer, Benjamin; Krajewski, Andrea; Coenen, Claudius; Siegmund, Dirk; Braun, Andreas

An Intuitive and Personal Projection Interface for Enhanced Self-management

2018

Distributed, Ambient, and Pervasive Interactions: Technologies and Contexts

International Conference on Distributed, Ambient and Pervasive Interactions (DAPI) <6, 2018, Las Vegas, NV, USA>

Smart environments offer a high potential to improve intuitive and personal interactions in our everyday life. Nowadays, we often get distracted by interfaces and have to adapt ourselves to the technology, instead of the interfaces focusing on the human needs. Especially in work situations, it is important to focus on the essential in terms of goal setting and to have a far-reaching vision about ourselves. Particularly with regard to self-employment, challenges like efficient self-management, regulated work times and sufficient self-reflection arise. Therefore, we present ‘Selv’, a novel transportable device that is intended to increase user productivity and self-reflection by having an overview about obligations, targets and success. ‘Selv’ is an adaptive interface that changes its interactions in order to fit into the user’s everyday routine. Our approach is using a pen on a projected interface. Adapting to our own feeling of naturalness ‘Selv’ learns usual interactions through handwriting recognition. In order to address users needs, it is more likely to built a mutual relationship and to convey a new feeling of an interface in a personal and natural way. This paper includes an elaborate concept and prototypical realization within the internet of things environment. We conclude with an evaluation of testings and improvements in terms of interactions and hardware.

Show publication details
Diaz-Rodriguez, Natalia; Grönroos, Stefan; Wickström, Frank; Lilius, Johan; Eertink, Henk; Braun, Andreas; Dillen, Paul; Crowley, James L.; Alexandersson, Jan

An Ontology for Wearables Data Interoperability and Ambient Assisted Living Application Development

2018

Recent Developments and the New Direction in Soft-Computing Foundations and Applications

World Conference on Soft Computing <6, 2016, Berkeley, USA>

Studies in Fuzziness and Soft Computing (STUDFUZZ)
361

Over the last decade a number of technologies have been developed that support individuals in keeping themselves active. This can be done via e-coaching mechanisms and by installing more advanced technologies in their homes. The objective of the Active Healthy Ageing (AHA) Platform is to integrate existing tools, hardware, and software that assist individuals in improving and/or maintaining a healthy lifestyle. This architecture is realized by integrating several hardware/software components that generate various types of data. Some examples include heart-rate data, coaching information, in-home activity patterns, mobility patterns, and so on. Various subsystems in the AHA platform can share their data in a semantic and interoperable way, through the use of a AHA data-store and a wearable devices ontology. This paper presents such an ontology for wearable data interoperability in Ambient Assisted Living environments. The ontology includes concepts such as height, weight, locations, activities, activity levels, activity energy expenditure, heart rate, or stress levels, among others. The purpose is serving application development in Ambient Intelligence scenarios ranging from activity monitoring and smart homes to active healthy ageing or lifestyle profiling.

  • 978-3-319-75407-9
Show publication details
Richter, Bernd; Urban, Bodo [Gutachter]; Schwender, Clemens [Zweitgutachter]

Animierte und interaktive Visualisierung in der Technischen Dokumentation

2018

Rostock, Univ., Master Thesis, 2018

Auf der tekom-Jahrestagung 2017 wurde im Schwerpunkt "Technical Videos" ein Vortrag gehalten mit dem Titel "Video is the new normal". Ein auf den ersten Blick schlichter Satz, dessen einfache Formulierung aber nicht über seine inhaltliche Signifikanz für gegenwärtige digitale Entwicklungen hinweg täuschen sollte. Schlägt man bei Oxford Dictionaries zur Bedeutung der Formulierung "the new normal" nach, so findet man: "A previously unfamiliar or atypical situation that has become standard, usual, or expected."1 Die Entwicklungen in den letzten Jahren hinsichtlich des verstärkten Konsums von Bewegtbild-Inhalten über das Internet bestätigen genau das: Das Abrufen von Videos über das Internet scheint in der Tat eine fast schon selbstverständliche Art der Informationsbeschaffung geworden zu sein. Alleine das Vorhandensein eines Schwerpunkts "Technical Videos" auf der Jahrestagung des größten europäischen Berufsverbandes für Technische Dokumentation, der tekom, deutet den Stellenwert an, den Bewegtbild-Medien mittlerweile innerhalb der Gesellschaft inne zu haben scheinen. Weitere Daten bestätigen Signifikanz und Beliebtheit dynamischer Medien: So ist die Videoplattform YouTube nach google die zweitgrößte Suchmaschine im Internet2 und verzeichnete im Juli 2015 einen Upload von mehr als 400 Stunden Video- Inhalte pro Minute3. Die große Popularität von Videos macht sie letzten Endes auch für die Technische Dokumentation relevant. In einer repräsentativen Umfrage des Digitalverbands Bitkom aus dem Jahre 20154 wurde ermittelt, dass mehr als ein Drittel (37%) der Internetnutzer ab 14 Jahren sich bereits Anleitungsvideos angesehen hat. Die so genannten "Tutorials" oder "how-to"-Videos erfreuen sich wachsender Beliebtheit und es ist anzunehmen, dass Nutzer, die sich im Alltag ständig über dynamische Medien wie Videos Informationen und Hilfestellungen besorgen, diese Form der Informationsvermittlung zukünftig auch verstärkt von den Erzeugnissen jener Berufsgruppe erwarten, deren Aufgabe es ist, Informationen und Anleitungen zu Produkten zu verfassen und bereitzustellen - der Technischen Dokumentation. Da nun die Erstellung eines Videos oder vergleichbaren Mediums tendenziell mit einem hohen konzeptionellen und operativen Aufwand verbunden ist, stellt sich für Technische Redakteure unweigerlich die Frage, ob und welchen Mehrwert dynamische Medien wie Videos speziell im instruktionalen Bereich im Vergleich zu gängigen textuellen oder aus Text-Bild-Kombinationen bestehenden Anleitungsmedien bieten können. Die vorliegende Arbeit möchte zur Klärung dieser Fragen beitragen. Sie beschäftigt sich mit der Wirksamkeit der so genannten kontinuierlichen Medien, speziell im Hinblick auf ihren Einsatz in der Technischen Dokumentation.

Show publication details
Kabir, Ahmedul; Kuijper, Arjan [1. Prüfer]; Kügler, David [2. Prüfer]

Annotation of X-ray Images in the Cloud: Analysis and Comparison of Perception Limits and Pose Estimation of Crowd Workers

2018

Darmstadt, TU, Master Thesis, 2018

The demand of medical image is increasing with every day that contains crucial information for diagnosis, treatment planning, disease monitoring, image-guided surgery, educating medical students through different medical cases and for many other research purposes in medical science. This information can be gathered through image classifications, registration, and segmentation. Machine learning algorithms are used for gathering information through medical image classification, registration or segmentation. However, these algorithms need a large set of training data in their initial stages to learn from. Medical experts provides the annotation of these medical images. This is where the problem lies. Annotating these medical images is a very time consuming, monotonous and expensive process. The medical experts lack motivation and are always occupied with their daily important clinician stuff. Thus, we need a solution that will be fast, accurate and cost-effective. This is where crowdsourcing comes into play. Yes, crowdsourcing is the best way to speed up these annotations tasks. But there are still questions whether these crowd workers are good enough to generate the initial training data sets for these algorithms and whether they are good enough to replace the experts. Recent research includes the detection of Malaria[LO12], detection of clinical features from Glaucomatous Optic Neuropathy [DMTP15] and Medical x-ray classification[LO12] where the crowd workers managed to perform as good as experts. The scope of our thesis is to analyze and compare the perceived limits and pose estimation of crowd workers for the annotation of x-ray images. To our prior knowledge, previously there has been no research done on this. We have the x-ray images of different parts of the body that contains bone surgeries with a screw in it. We have put two wireframed screws, one in red and the other in green color and put both beside the ground-truth screw. And the objective of this thesis is to study whether these crowds are able to interpret the x-ray image and then able to classify which screw is close to the original ground-truth screw. We ran the first x-ray classification experiment with our Master students as a controlled group. And then we ran the next two experiments in Amazon Mechanical Turk(AMT). When the mTurk workers were selected with an approval rate over 90% and 500 HITs completion, they underperformed. The accuracy rate was below 45% for Euclidean distance difference between the screws from 1px to 3px. We repeated the experiment with mTurk workers with a higher approval rate over 98% and HITs completion over 1,000. We could immediately see the mTurk workers performed well We could see when the difference in Euclidean distance between the screws is 1 pixel or less, both the controlled and the mTurk workers could classify with an accuracy of 80%. Therefore, we can easily say crowd workers can replace medical experts for generating training data for algorithms in medical x-ray image classifications with a perception limit of 1px with respect to Euclidean distance. However, we failed to do our second experiment which was about post estimation of crowd workers for the registration tasks in the x-ray images. We managed to build the prototype but we failed to map the projection data of the ground truth screw with our model screw. But the prototype can be used in future to run the experiment once the projection matrix could be mapped out.

Show publication details
Damer, Naser; Kuijper, Arjan [Referent]; Fellner, Dieter W. [Referent]; Ramachandra, Raghavendra [Referent]

Application-driven Advances in Multi-biometric Fusion

2018

Darmstadt, TU, Diss., 2018

Biometric recognition is the automated recognition of individuals based on their behavioral or biological characteristics. Beside forensic applications, this technology aims at replacing the outdated and attack prone, physical and knowledge-based, proofs of identity. Choosing one biometric characteristic is a tradeoff between universality, acceptability, and permanence, among other factors. Moreover, the accuracy cap of the chosen characteristic may limit the scalability and usability for some applications. The use of multiple biometric sources within a unified frame, i.e. multi-biometrics, aspires to tackle the limitations of single source biometrics and thus enables a wider implementation of the technology. This work aims at presenting application-driven advances in multi-biometrics by addressing different elements of the multi-biometric system work-flow. At first, practical oriented pre-fusion issues regarding missing data imputation and score normalization are discussed. This includes presenting a novel performance anchored score normalization technique that aligns certain performance-related score values in the fused biometric sources leading to more accurate multi-biometric decisions when compared to conventional normalization approaches. Missing data imputation within scorelevel multi-biometric fusion is also addressed by analyzing the behavior of different approaches under different operational scenarios. Within the multi-biometric fusion process, different information sources can have different degrees of reliability. This is usually influenced in the fusion process by assigning relative weights to the fused sources. This work presents a number of weighting approaches aiming at optimizing the decision made by the multi-biometric system. First, weights that try to capture the overall performance of the biometric source, as well as an indication of its confidence, are proposed and proved to outperform the state-of-the-art weighting approaches. The work also introduces a set of weights derived from the identification performance representation, the cumulative match characteristics. The effect of these weights is analyzed under the verification and identification scenarios. To further optimize the multi-biometric process, information besides the similarity between two biometric captures can be considered. Previously, the quality measures of biometric captures were successfully integrated, which requires accessing and processing raw captures. In this work, supplementary information that can be reasoned from the comparison scores are in focus. First, the relative relation between different biometric comparisons is discussed and integrated in the fusion process resulting in a large reduction in the error rates. Secondly, the coherence between scores of multi-biometric sources in the same comparison is defined and integrated into the fusion process leading to a reduction in the error rates, especially when processing noisy data. Large-scale biometric deployments are faced by the huge computational costs of running biometric searches and duplicate enrollment checks. Data indexing can limit the search domain leading to faster searches. Multibiometrics provides richer information that can enhance the retrieval performance. This work provides an optimizable and configurable multi-biometric data retrieval solution that combines and enhances the robustness of rank-level solutions and the performance of feature-level solutions. Furthermore, this work presents biometric solutions that complement and utilize multi-biometric fusion. The first solution captures behavioral and physical biometric characteristics to assure a continuous user authentication. Later, the practical use of presentation attack detection is discussed by investigating the more realistic scenario of cross-database evaluation and presenting a state-of-the-art performance comparison. Finally, the use of multibiometric fusion to create face references from videos is addressed. Face selection, feature-level fusion, and score-level fusion approaches are evaluated under the scenario of face recognition in videos.

Show publication details

Augmented Reality Views: Discussing the Utility of Visual Elements by Mediation Means in Industrial AR from a Design Perspective

2018

Virtual Augmented and Mixed Reality: Applications in Health, Cultural Heritage, and Industry

International Conference Virtual Augmented and Mixed Reality (VAMR) <10, 2018, Las Vegas, NV, USA>

Lecture Notes in Computer Science (LNCS)
10910

In this paper we present and discuss common visual elements in Augmented Reality which create a distinct information context while presenting them in either video- or optical-see-though setups, and which align with the promise of AR being able to bridge the gap between real world objects and the digital information space about them. Reflecting on nowadays common elements with these premises in mind, we collected and categorized a variety of visual elements, e.g. annotation & labels, visual highlights, assisting visual aids and trans-media elements. Focusing on industrial AR applications, we discuss their suitability in terms of mediation and communication goals, instead of technological and implementation considerations. In doing so, we seek to identify the currently most relevant visual elements and discuss the deployed meaning that can be created in utilizing these elements for a informed and successful communication. From there we introduce a first framing meta-model that on the one hand helps clarifying the mediation strength of these elements and on the other enables to reflect their suitability on a more strategic level.

  • 978-3-319-91583-8
Show publication details

Automated Acquisition and Real-time Rendering of Spatially Varying Optical Material Behavior

2018

ACM SIGGRAPH 2018 Posters

International Conference on Computer Graphics and Interactive Techniques (SIGGRAPH) <45, 2018, Vancouver, BC, Canada>

We created a fully automatic system for acquisition of spatially varying optical material behavior of real object surfaces under a hemisphere of individual incident light directions. The resulting measured material model is flexibly applicable to arbitrary 3D model geometries, can be photorealistically rendered and interacted with in real-time and is not constrained to isotropic materials.

Show publication details
Judzinsky, Nils; Lukas, Uwe von [Gutachter]; Urban, Bodo [Zweitgutachter]; Albadawi, Mohamad [Betreuer]; Krause, Tom [Betreuer]

Automated Object Pose Recognition by a Combination of Stereo Cameras and 2D Object Detection

2018

Rostock, Univ., Master Thesis, 2018

The ability to visually recognize the 3D pose of objects would be helpful in many industrial applications. Existing approaches on this topic restrict the pose estimation to simplified scenarios, e.g. where a pose consists of just a 2D position and an angle around the vertical axis, or they require a priori knowledge. In this thesis methods are investigated to estimate a 6 DoF pose and 3D extends of arbitrary objects captured by a stereo camera. No knowledge about the shape of objects is required beforehand. A prototype implementation is provided. It first detects the objects in 2D employing a CNN trained for object detection. Then, the depth information is used to reconstruct a 3D point cloud and isolate the objects of interest. Finally, pose and size are estimated based on dense and sparse registration methods. At the end, the whole method is tested on artificially generated stereo images of fish. The results show remaining challenges especially regarding the robustness.

Show publication details
Deshmukh, Akshay Madhav; Kuijper, Arjan [1. Gutachten]; Burkhardt, Dirk [2. Supervisor]

Automated User Evaluation Analysis for a Simplified and Continuous Software Development

2018

Darmstadt, TU, Master Thesis, 2018

In today's world, computers are tightly coupled with the internet and play a vital role in the development of business and various aspects of human lives. Hence, developing a quality user-computer interface has become a major challenge. Well-designed programs that are easily usable by users are moulded through a regress development life cycle. To ensure a user friendly interface, the interface has to be well designed and need to support smart interaction features. User interface can become an Archilles heel in a developed system because of the simple design mistakes which causes critical interaction problems which eventually leads to massive loss of attractiveness in the system. To overcome this problem, regular and consistent user evaluations have to be carried out to ensure the usability of the system. The importance of an evaluation for the development of a system is well known. Most of the today's existing approaches necessitate the users to carry out an evaluation in a laboratory. Evaluators are compelled to dedicate the time in informing the participants about the evaluation process and providing a clear understanding of the questionnaires during the experiment. At the post experiment phase, evaluators have to invest a huge amount of time in generating a result report. On the whole, most of the today's existing evaluation approaches hogs up too much of time for most developments. The main aim of this thesis is to develop an automated evaluation management and result analysis, based on a previous developed web-based evaluation system, which enables to elaborate the evaluation results and identify required changes on the developed system. The major idea is that an evaluation can be prepared once and repeated in regular time intervals with different user groups. The automated evaluation result analysis allows to easily check if the continued development lead to better results and if a bunch of given task could be better solved e.g. by added new functions or through enhanced presentation. Within the scope of this work, Human-Computer Interaction (HCI) was researched, in particular towards User-Centered Design (UCD) and User Evaluation. Different approaches for an evaluation were researched in particular towards an evaluation through expert analysis and user participation. Existing evaluation strategies and solutions, inclined towards distributed evaluations in the form of practical as well as survey based evaluation methods were researched. A proof of concept of an automated evaluation result analysis that enables an easy detection of gaps and improvements in the system was implemented. Finally, the results of the research project Smarter Privacy were compared with the manual performed evaluation.

Show publication details
Limper, Max; Fellner, Dieter W. [Referent]; Alexa, Marc [Referent]

Automatic Optimization of 3D Mesh Data for Real-Time Online Presentation

2018

Darmstadt, TU, Diss., 2018

Interactive 3D experiences are becoming increasingly available as a part of our every-day life. Examples are ranging from common video games to virtual reality experiences and augmented reality apps on smart phones. A rapidly growing area are interactive 3D applications running inside common Web browsers, enabling to serve millions of users worldwide using solely standard Web technology. However, while Web-based 3D presentation technology is getting more and more advanced, a crucial problem that remains is the optimization of 3D mesh data, such as highly detailed 3D scans, for efficient transmission and online presentation. In this context, the need for dedicated 3D experts, being able to work with various specialized tools, significantly limits the scalability of 3D optimization workflows in many important areas, such as Web-based 3D retail or online presentation of cultural heritage. Moreover, since Web-based 3D experiences are nowadays ubiquitous, an optimal delivery format must work well on a wide range of possible client devices, including tablet PCs and smart phones, while still offering acceptable compression rates and progressive streaming. Automatically turning high-resolution 3D meshes into compact 3D representations for online presentations, using an efficient standard format for compression and transmission, is therefore an important key challenge, which remained largely unsolved so far. Within this thesis, a fully-automated pipeline for appearance-preserving optimization of 3D mesh data is presented, enabling direct conversion of high-resolution 3D meshes to an optimized format for real-time online presentation. The first part of this thesis discusses 3D mesh processing algorithms for fully-automatic optimization of 3D mesh data, including mesh simplification and texture mapping. In this context, a novel saliency detection method for mesh simplification is presented, as well as a new method for automatic overlap removal in parameterizations using cuts with minimum length and, finally, a method to compact texture atlases using a cut-and-repack strategy. The second part of the thesis deals with the design of an optimized format for 3D mesh data on the Web. It covers various relevant aspects, such as efficient encoding of mesh geometry and mesh topology, a physically-based format for material data, and progressive streaming of textured triangle meshes. The contributions made in this context during the creation of this thesis had notable impact on the design of the current standard format for 3D mesh data on the Web, glTF 2.0, which is nowadays supported by the vast majority of online 3D viewers.

Show publication details

Automatic Procedural Model Generation for 3D Object Variation

2018

The Visual Computer

3D objects are used for numerous applications. In many cases not only single objects but also variations of objects are needed. Procedural models can be represented in many different forms, but generally excel in content generation. Therefore this representation is well suited for variation generation of 3D objects. However, the creation of a procedural model can be time-consuming on its own. We propose an automatic generation of a procedural model from a single exemplary 3D object. The procedural model consists of a sequence of parameterizable procedures and represents the object construction process. Changing the parameters of the procedures changes the surface of the 3D object. By linking the surface of the procedural model to the original object surface, we can transfer the changes and enable the possibility of generating variations of the original 3D object. The user can adapt the derived procedural model to easily and intuitively generate variations of the original object. We allow the user to define variation parameters within the procedures to guide a process of generating random variations. We evaluate our approach by computing procedural models for various object types, and we generate variations of all objects using the automatically generated procedural model.

Show publication details
Wirtz, Andreas; Wambach, Johannes; Wesarg, Stefan

Automatic Teeth Segmentation in Cephalometric X-Ray Images Using a Coupled Shape Model

2018

OR 2.0 Context-Aware Operating Theaters, Computer Assisted Robotic Endoscopy, Clinical Image-Based Procedures, and Skin Image Analysis

International Workshop on OR 2.0 Context-Aware Operating Theaters (OR 2.0) <1, 2018, Granada, Spain>

Lecture Notes in Computer Science (LNCS)
11041

Cephalometric analysis is an important tool used by dentists for diagnosis and treatment of patients. Tools that could automate this time consuming task would be of great assistance. In order to provide the dentist with such tools, a robust and accurate identification of the necessary landmarks is required. However, poor image quality of lateral cephalograms like low contrast or noise as well as duplicate structures resulting from the way these images are acquired make this task difficult. In this paper, a fully automatic approach for teeth segmentation is presented that aims to support the identification of dental landmarks. A 2-D coupled shape model is used to capture the statistical knowledge about the teeth’s shape variation and spatial relation to enable a robust segmentation despite poor image quality. 14 individual teeth are segmented and labeled using gradient image features and the quality of the generated results is compared to manually created gold-standard segmentations. Experimental results on a set of 14 test images show promising results with a DICE overlap of 77.2% and precision and recall values of 82.3% and 75.4%, respectively.

  • 978-3-030-01200-7
Show publication details
Wirtz, Andreas; Mirashi, Sudesh Ganapati; Wesarg, Stefan

Automatic Teeth Segmentation in Panoramic X-Ray Images Using a Coupled Shape Model in Combination with a Neural Network

2018

Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: Part IV

International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) <21, 2018, Granada, Spain>

Lecture Notes in Computer Science (LNCS)
11073

Dental panoramic radiographs depict the full set of teeth in a single image and are used by dentists as a popular first tool for diagnosis. In order to provide the dentist with automatic diagnostic support, a robust and accurate segmentation of the individual teeth is required. However, poor image quality of panoramic x-ray images like low contrast or noise as well as teeth variations in between patients make this task difficult. In this paper, a fully automatic approach is presented that uses a coupled shape model in conjunction with a neural network to overcome these challenges. The network provides a preliminary segmentation of the teeth region which is used to initialize the coupled shape model in terms of position and scale. Then the 28 individual teeth (excluding wisdom teeth) are segmented and labeled using gradient image features in combination with the model’s statistical knowledge about their shape variation and spatial relation. The segmentation quality of the approach is assessed by comparing the generated results to manually created goldstandard segmentations of the individual teeth. Experimental results on a set of 14 test images show average precision and recall values of 0.790 and 0.827, respectively and a DICE overlap of 0.744.

  • 978-3-030-00936-6
Show publication details
Hofmann, Patrick; Bredel, Michael [Referent]; Stiemerling, Martin [Korreferent]; Oyarzun Laura, Cristina [Betreuerin]

Automatische Erkennung der Nasennebenhöhlen mit Hilfe eines neuronalen Netzes

2018

Darmstadt, Hochschule, Bachelor Thesis, 2018

In der Medizin ist es für verschiedene Eingriffe wichtig, die Strukturen im Schädel zu erkennen, um ggf. empfindliche Strukturen bei einem Eingriff nicht in Mitleidenschaft zu ziehen. Mittels bildgebender Diagnostik können so z.B. CT-Bilder entstehen bei denen ein Arzt oder eine Fachkraft dann per Hand die Strukturen eintragen (markieren) kann. Da dieser Prozess sehr langwierig sein kann und sich über mehrere Stunden hinziehen kann, werden Lösungen gesucht die Erkennung zu automatisieren. Bei knöchernen oder gleichbleibenden Strukturen ist dies bereits zufriedenstellend gelungen, jedoch sind die Nebenhöhlen von Patient zu Patient stark verschieden und können mit den bestehenden Verfahren nicht mit der ausreichenden Präzision umgesetzt werden. So dass die Nacharbeit inkl. der Überprüfung dem Fachpersonal keine wirkliche Zeitersparnis bringt. In dieser Arbeit wird mit der Hilfe eines neuronalen Netzes versucht die automatische Erkennung der Nebenhöhlen (bestehend aus der Stirnhöhle, der Keilbeinhöhle, rechter und linker Kieferhöhle, rechter und linker Siebbeinzelle und der Nasenhöhle) zu automatisieren. Dabei liegt die Herausforderung nicht nur an der starken Varianz der zu erkennenden Struktur, sondern auch darin, dass nur eine für neuronale Netze vergleichbare geringe Zahl an Lerndaten vorhanden ist. Wider dem Lerndatenmangel wird versucht durch Betrachtung der einzelnen 2D-Schichten statt des kompletten 3D-Datensatzes zu umgehen. So wird in der Arbeit mittels DarkFlow ein YoloV2 Netz trainiert und die Ergebnisse in der Validierung diskutiert.

Show publication details
Puhl, Julian; Kuijper, Arjan [1. Gutachten]; Jung, Florian [2. Gutachten]

Automatische Tumordetektion und Segmentierung im Bereich der Mundhöhle anhand der Symmetrie des menschlichen Kopfes

2018

Darmstadt, TU, Master Thesis, 2018

Die Diagnose von Tumoren im Kopf-Hals-Bereich bedeuten für den Patienten in der Regel eine große Einschränkung der Lebensqualität. Aus diesem Grund wird versucht, die Behandlung sehr genau auf den jeweiligen Patienten abzustimmen. Ein wichtiges Hilfsmittel dabei sind Segmentierungen der Tumore auf aufgenommenen Bilddaten. Für die Aufnahme kommt oft die Magnetresonanztomographie zum Einsatz. Für diesen speziellen Anwendungsfall existiert bisher kein automatisches Verfahren, welches gute Ergebnisse liefert. An diesem Punkt setzt diese Arbeit an. Es wird ein Verfahren vorgestellt, welches auf MRT-Aufnahmen automatisch Tumore erkennt und segmentiert. Voraussetzung ist, dass die Aufnahme mit Verfahren zur Unterdrückung von Fettgewebe, wie z.B. SPIR, erstellt wird und die Gewichtung der Aufnahme T1 ist. Außerdem muss dem Patienten Kontrastmittel zur Hervorhebung des Tumorgewebes verabreicht worden sein. Das Verfahren nutzt die Symmetrie des menschlichen Kopfes, um Asymmetrie und somit mögliche Tumore in der Aufnahme zu finden. Hierfür wird die Symmetrieachse mittels Optimierungsverfahren und geeigneten Parametern bestimmt. Zusätzlich wird als weiteres Merkmal das durch Kontrastmittel in der Aufnahme aufgehellte Gewebe verwendet. Diese beiden Merkmale kombiniert ergeben als Zwischenergebnis eine Maske, von der die drei größten zusammenhängenden Komponenten extrahiert werden. Diese werden als Basis für eine weitere Segmentierung verwendet. Hierbei kommt das Bereichswachstumsverfahren (engl. Region Growing) mit in dieser Arbeit entworfenem Kriterium zum Einsatz. Als Ergebnis erhält man bis zu drei Segmentierungen möglicher Tumorkandidaten, von denen der Arzt entscheiden kann, welche er verwenden möchte. Evaluiert wurde das Verfahren auf 40 MRT-Aufnahmen, welche mit T1-Gewichtung, dem SPIR Verfahren zur Fettunterdrückung und verabreichtem Kontrastmittel aufgenommen wurden. Es standen hierbei für alle Aufnahmen von Ärzten erstellte Segmentierungen der Tumore zur Verfügung. Als Ergebnis wird ein Dice-Durchschnittswert von 0,64 und eine durchschnittliche Hausdorff-Distanz von 21,48 erzielt. Die Erkennungsrate liegt dabei bei sehr guten 90%. Der Vorteil des Verfahrens ist, dass es vollautomatisch funktioniert und potentiell sogar mehrere Tumore finden könnte. Für die Endauswahl der Segmentierungen kann Expertenwissen einfließen und bspw. können zusätzlich die gewählten Segmentierungen manuell weiter verfeinert werden. Dies bedeutet im Normalfall eine große Zeitersparnis, da bei heutigem Arbeitsablauf die Segmentierungen oft noch vollständig per Hand erstellt werden.

Show publication details
Limper, Max; Vining, Nicolas; Sheffer, Alla

Box Cutter: Atlas Refinement for Efficient Packing via Void Elimination

2018

ACM Transactions on Graphics

Packed atlases, consisting of 2D parameterized charts, are ubiquitously used to store surface signals such as texture or normals. Tight packing is similarly used to arrange and cut-out 2D panels for fabrication from sheet materials. Packing efficiency, or the ratio between the areas of the packed atlas and its bounding box, significantly impacts downstream applications. We propose Box Cutter, a new method for optimizing packing efficiency suitable for both settings. Our algorithm improves packing efficiency without changing distortion by strategically cutting and repacking the atlas charts or panels. It preserves the local mapping between the 3D surface and the atlas charts and retains global mapping continuity across the newly formed cuts. We balance packing efficiency improvement against increase in chart boundary length and enable users to directly control the acceptable amount of boundary elongation. While the problem we address is NP-hard, we provide an effective practical solution by iteratively detecting large rectangular empty spaces, or void boxes, in the current atlas packing and eliminating them by first refining the atlas using strategically placed axis-aligned cuts and then repacking the refined charts. We repeat this process until no further improvement is possible, or until the desired balance between packing improvement and boundary elongation is achieved. Packed chart atlases are only useful for the applications we address if their charts are overlap-free; yet many popular parameterization methods, used as-is, produce atlases with global overlaps. Our pre-processing step eliminates all input overlaps while explicitly minimizing the boundary length of the resulting overlap-free charts. We demonstrate our combined strategy on a large range of input atlases produced by diverse parameterization methods, as well as on multiple sets of 2D fabrication panels. Our framework dramatically improves the output packing efficiency on all inputs; for instance with boundary length increase capped at 50% we improve packing efficiency by 68% on average.

Show publication details

CAE/VR Integration – A Qualitative Assessment of Advanced Visualization for Interactive Conceptual Simulations (ICS) in Industrial Use

2018

Virtual Augmented and Mixed Reality: Applications in Health, Cultural Heritage, and Industry

International Conference Virtual Augmented and Mixed Reality (VAMR) <10, 2018, Las Vegas, NV, USA>

Lecture Notes in Computer Science (LNCS)
10910

One of the key driving technologies for a better communication, representation, interaction, and visualization of design and engineering data has been Virtual Reality (VR). The idea of Interactive Conceptual Simulations (ICS) combines real-time interaction and visualization in a turn-around loop with the CAE simulation. However, the automation of the processes between the changes of the domain and the resulting simulation requires update rates within min 30 Hz, in order to remain interactive, not blocking the CG engine redrawing the update. Here, the overall simulation requires an advanced CAE process chain from model import, model manipulation, simulation results generation and visual preparation. Although many publications have been addressing the CAD/VR and only few the CAE/VR process chain, one might assume that endeavors for this are regarded as unnecessary, as a tidy showcase with a low-level quality visualization would be sufficient for engineers. This paper will object this hypothesis by presenting the results of a qualitative validation based on industrial use of our established VR environment for ICS. It follows the methodology and complements the quantitative analysis presented in earlier work providing the results of a qualitative assessment of engineers for our established VR based ICS post-processing unit (IDEFix – Immersive Data Explorer for CAx Integration).

  • 978-3-319-91583-8
Show publication details
Ivanov, Ivelin; Kuijper, Arjan [Advisor]; Wilmsdorff, Julian von [1. Supervisor]; Kirchbuchner, Florian [2.. Supervisor]

CapBed - Preventive Assistance System for the Bed Area Based on Capacitive Sensing

2018

Darmstadt, TU, Master Thesis, 2018

Over the past decades, human activity recognition systems have become a major input modality for building automation. However, those systems also found recent applications in emergency detection, such as recognizing patient activities that may lead to life-threatening situations like falls or heart attacks. The aim of this thesis is to develop a sensor that recognizes whether a person wants to get out of bed. This is to prevent falls by illuminating the path or calling a nurse in time. In addition, such a system can also provide insights into the behavior of the user in the long term. Therefore, a concept of preventive assistance system for the bed area based on capacitive sensing is developed within the scope of this work. To this end, a comparison to other sensor technologies will be established, followed by a detailed overview of the technical background of capacitive proximity sensing. An innovative concept of a device that offers decent performance at an affordable price is proposed. Based on this concept a prototype system was developed and evaluated to investigate its sensing performance and identify possible limitations. As a future outlook, this thesis summarizes the occurred problems and suggests possible modifications that might improve the overall performance of the system.

Show publication details
Medina, Francisco; Kuijper, Arjan [Prüfer]; Fu, Biying [Betreuerin]

CapFloor - a Smart Floor for Sport Exercise Recognition

2018

Darmstadt, TU, Bachelor Thesis, 2018

Advances in sensor technology and computer systems have sparked interest in context-aware applications aimed to improve the quality of life of individuals. Motion detection, localization, tracking and, most importantly, interpretation of human behavior have been the focus in the development of smart environments. While vision-based systems and wearables solutions lead in terms of maturity, there is growing interest in low-cost, non-intrusive and privacy-preserving technologies. In recent years machine learning has rapidly been adopted in many industries and research fields. Computer vision, audio processing and recommendation engines are some examples that have greatly benefited from data-driven prediction models. Advances in these areas as well as the rapid adoption of Internet of Things (IoT) devices have enabled new and more reliable ways for smart environments to improve the day-to-day life of individuals. This thesis contributes a machine learning approach to sport activity recognition using passive electric field sensing. In this work multiple artificial neural networks are explored for the task of classifying 8 distinct sport activities, borrowing from techniques used in well established areas such as human activity recognition, image and video classification. The models explored are a 3D convolutional neural network (CNN) using only data from the passive electric field sensing system, a long short-term memory (LSTM) recurrent neural network (RNN) with accelerometer data for comparison, and finally an artificial neural network (ANN) composed of the two former models. The proposed system provides a low-cost, non-intrusive smart floor solution with a variety of use-cases ranging from fitness studios, sport rehabilitation centers and smart homes.

Show publication details
Jarms, Lennart; Kuijper, Arjan [Prüfer]; Fu, Biying [Betreuerin]

CapMat for Sport Exercise Recognition and Tracking

2018

Darmstadt, TU, Master Thesis, 2018

A large variety of physical exercises can be performed on the ground, typically using a dedicated mat. Many of these exercises do not require additional equipment and mainly consist of specific movements of different body parts. Monitoring sport exercises, i.e. recognition, tracking and counting, has been well researched to help motivate regular exercise and aid in physical rehabilitation. Most of the suggested systems rely on wearable devices or smartphones, which are not always at hand and depend on the limb they are attached to. Camera based solutions are usually not portable and raise privacy concerns.Using dedicated pressure mats has shown great success, but is limited in their adaption to online applications due to their costly computation caused by high sensor resolution. While a few prototypes have been suggested, there is no commercially available product yet, suggesting the difficulty of this area and the need for further research. Furthermore, it is restricted to exercises which are distinguishable by changing the contact area or weight distribution on the mat.We introduce CapMat, a smart sports mat that reaches a user independent recognition rate of 93.5 % in a user study with 9 subjects performing 8 exercises. It is developed with the Open Cap Sense (OCS) board with 8 copper plates as electrodes for capacitive proximity sensing hidden beneath a common sports mat. Moreover, we demonstrate its ability to count exercise repetitions, achieving 93.8 % repetition recognition rate for 12 exercise sets from a single user. The thesis focuses on robust activity classification and several methods to reach this objective are discussed, such as model selection, feature selection and data augmentation.The goal is to develop a system using capacitive sensing to recognize a wide range of exercises. The process from choosing electrode materials to their placement beneath the mat is discussed. The system allows the usage in real-time applications, which is demonstrated with a simple web application running on a Raspberry Pi 3.

Show publication details
Gödde, Michael; Gabler, Frank; Siegmund, Dirk; Braun, Andreas

Cinematic Narration in VR – Rethinking Film Conventions for 360 Degrees

2018

Virtual Augmented and Mixed Reality: Applications in Health, Cultural Heritage, and Industry

International Conference Virtual Augmented and Mixed Reality (VAMR) <10, 2018, Las Vegas, NV, USA>

The rapid development of VR technology in the past three years allowed artists, filmmakers and other media producers to create great experiences in this new medium. But filmmakers are, however, facing big challenges, when it comes to cinematic narration in VR. The old, established rules of filmmaking do not apply for VR films and important techniques of cinematography and editing must be completely rethought. Possibly, a new filmic language will be found. But even though filmmakers eagerly experiment with the new medium already, there exist relatively few scientific studies about the differences between classical filmmaking and filmmaking in 360 and VR. We therefore present this study on cinematic narration in VR. In this we give a comprehensive overview of techniques and concepts that are applied in current VR films and games. We place previous research on narration, film, games and human perception into the context of VR experiences and we deduce consequences for cinematic narration in VR. We base our assumptions on a conducted empirical test with 50 participants and on an additional online survey. In the empirical study, we selected 360-degree videos and showed them to a test-group, while the viewer’s behavior and attention was observed and documented. As a result of this paper, we present guidelines which suggest methods of guiding the viewers’ attention as well as approaches to cinematography, staging and editing in VR.

Show publication details
Bernard, Jürgen; Hutter, Marco; Zeppelzauer, Matthias; Fellner, Dieter W.; Sedlmair, Michael

Comparing Visual-Interactive Labeling with Active Learning: An Experimental Study

2018

IEEE Transactions on Visualization and Computer Graphics

Labeling data instances is an important task in machine learning and visual analytics. Both fields provide a broad set of labeling strategies, whereby machine learning (and in particular active learning) follows a rather model-centered approach and visual analytics employs rather user-centered approaches (visual-interactive labeling). Both approaches have individual strengths and weaknesses. In this work, we conduct an experiment with three parts to assess and compare the performance of these different labeling strategies. In our study, we (1) identify different visual labeling strategies for user-centered labeling, (2) investigate strengths and weaknesses of labeling strategies for different labeling tasks and task complexities, and (3) shed light on the effect of using different visual encodings to guide the visual-interactive labeling process. We further compare labeling of single versus multiple instances at a time, and quantify the impact on efficiency. We systematically compare the performance of visual interactive labeling with that of active learning. Our main findings are that visual-interactive labeling can outperform active learning, given the condition that dimension reduction separates well the class distributions. Moreover, using dimension reduction in combination with additional visual encodings that expose the internal state of the learning model turns out to improve the performance of visual-interactive labeling.

Show publication details

Computing Feedback for Citizens’ Proposals in Participative Urban Planning

2018

3rd International Conference on Smart Data and Smart Cities

International Conference on Smart Data and Smart Cities <3, 2018, Delft, The Netherlands>

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
IV-4/W7

We show an approach how to provide computed feedback on citizens’ proposals based on open data and expert knowledge in urban planning and public participation by using Domain-Specific Languages (DSL). We outline the process involving different stakeholders of engineering such a DSL and provide an architecture capable of executing the language and uploading new scripts at runtime. A real-world example of the city of Hamburg is used to show the principles and serves as input for development. A prototype has been implemented and evaluated at various events involving citizen and city representatives. We conclude that DSLs can be successfully applied to enable a new way to access data in a more convenient and understandable form, abstracting from technical details and focusing on domain aspects.

Show publication details

Continuous Property Gradation for Multi-material 3D-printed Objects

2018

Solid Freeform Fabrication 2018: Proceedings of the 29th Annual International Solid Freeform Fabrication Symposium - An Additive Manufacturing Conference

Annual International Solid Freeform Fabrication Symposium - An Additive Manufacturing Conference <29, 2018, Austin, TX, USA>

Modern AM processes allow for printing multiple materials. The resulting objects can be stiff/dense in some areas and soft/porous in others, resulting in distinct physical properties. However, modeling material gradients is still tedious with current approaches, especially when smooth transitions are required. Current approaches can be distinguished into a) NURBS-BReps-based and b) voxel-based. In case of NURBS-BReps, discrete material distributions can be modeled by manually introducing separate shells inside the object; smooth gradation can only be approximated in discrete steps. For voxel representations, gradation is discrete by design and comes along with an approximation error. In addition, interacting on a per-voxel basis is tedious for the designer/engineer. We present a novel approach for representing material gradients in volumetric models using subdivision schemes, supporting continuity and providing elegant ways for interactive modeling of locally varying properties. Additionally, the continuous volumetric representation allows for on-demand sampling at any resolution required by the 3D printer.

Show publication details

CPROSA-Holarchy: An Enhanced PROSA Model to Enable Worker–Cobot Agile Manufacturing

2018

International Journal of Mechanical Engineering and Robotics Research

This research combines two important concepts of intelligent manufacturing: agile manufacturing and collaborative robotics. On the one hand, agility in manufacturing is the capability of an industrial enterprise to respond rapidly and effectively to unanticipated changes that occur in the production. The aim of agile manufacturing is to proactively develop solutions to adapt to the customers’ needs. These solutions are a result of collective decision-making that has been formed among the different entities of the agile manufacturing system. On the other hand, collaborative robotics is a new trend in industrial robotics, which involves a collaborative robot (cobot). A cobot is usually an industrial robot designed to operate safely in a shared work environment with human workers. This is in contrast with the conventional industrial robot that operates in isolation from the workers’ workspace. One of the most important advantages of collaborative robotics is the increase of the agility of manufacturing. Therefore, in this research, we focus on developing a proper information control and communication solution to facilitate worker–cobot agile manufacturing. Furthermore, we introduce a case study of two workers in cooperation with one cobot to demonstrate the solution concept.

Show publication details
Damer, Naser; Wainakh, Yaza; Boller, Viola; Berken von den, Sven; Terhörst, Philipp; Braun, Andreas; Kuijper, Arjan

CrazyFaces: Unassisted Circumvention of Watchlist Face Identification

2018

IEEE 9th International Conference on Biometrics: Theory, Applications and Systems

IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS) <9, 2018, Redondo Beach, CA, USA>

Once upon a time, there was a blacklisted criminal who usually avoided appearing in public. He was surfing the Web, when he noticed, what had to be a targeted advertisement announcing a concert of his favorite band. The concert was in a near town, and the only way to get there was by train. He was worried, because he heard in the news about the new face identification system installed at the train station. From his last stay with the police, he remembers that they took these special face images with the white background. He thought about what can he do to avoid being identified and an idea popped in his mind “what if I can make a crazy-face, as the kids call it, to make my face look different? What do I exactly have to do? And will it work?”. He called his childhood geeky friend and asked him if he can build him a face recognition application he can tinker with. The geeky friend was always interested in such small projects where he can use open-source resources and didn’t really care about the goal, as usual. The criminal tested the application and played around, trying to figure out how can he make a crazy-face that won’t be identified as himself. On the day of the concert, he took off to the train station with some doubt in his mind and fear in his soul. To know what happened next, you should read the rest of this paper.

  • 978-1-5386-7180-1
Show publication details
Moseguí Saladié, Alexandra; Yildirim Yayilgan, Sule [Supervisor]; Damer, Naser [Supervisor]

Creating Face Morphing Attacks with Generative Adversarial Networks

2018

Saint-Étienne, Univ., Master Thesis, 2018

Nowadays, the use of technologies related to biometrics is increasing significantly. The recent deep learning improvement in face recognition tasks, along with the relatively high social acceptance, have pushed automatic face recognition systems to be a key technology in identity verification in border controls. In this scenario, face recognition is used to link the identity of a passenger to their e-document. However, recent studies have highlighted the threat of morphing attacks against automatic face recognition systems. In this thesis, we present a novel morphed face attack, called MorGANA, created by using Generative Adversarial Networks. By creating a new morphing database, MorGAN dataset, we investigate the vulnerabilities of current face recognition systems against MorGANA attacks, alongside with baseline attacks. Moreover, the morphing detectability under face morphing attacks is further studied noticing an insufficient performance from common detectors. Examining the obtained results, we strongly consider that further studies need to be addressed in generating and detecting morphed face attacks in order to bring face recognition systems to real-case scenarios.

Show publication details
Anvekar, Roshan; Damer, Naser [Supervisor]; Pech, Andreas [Supervisor]; Nauth, Peter [Supervisor]

Cross-device Biometric Verification and the Benefit of Multi-algorithmic Biometric Fusion

2018

Frankfurt am Main, Univ. of Applied Science, Master Thesis, 2018

Biometrics is a technology that aims to identify or verify people identities based on their physical characteristics or behavioral properties. Multi-biometrics is implemented to use a number of biometric information sources to create a unified decision to increase the accuracy of the biometric system. This aims at enhancing performance and avoiding the shortcomings of conventional single source biometrics such as sensitivity to noisy data or capture environment while maintaining high accuracy. One of these shortcomings are the negative effect of cross-device biometric verification, e.g. using different smart phones and capture devices for the enrolment and verification of face images. The previous work deals with multi-biometrics to obtain the fused score by incorporating the coherence information on the biometric sources to enhance the performance of the Biometric system. In this thesis the main purpose is to improve the performance of the Biometric system for multi-modality (two modalities) with each biometric source acquired from different sensor devices which includes both same sensor and different sensor combinations. This includes biometric sources, i.e. face and one of the periocular region fed to the biometric system with multi-algorithmic approach to extract the features and create unified decision at the score level. Analysis of different realistic scenario with varying static and dynamic weights and its effect on the unified decision on score level. This approach could be used in biometric system with more than two modalities.

Show publication details

CultArc3D_mini: Fully Automatic Zero-Button 3D Replicator

2018

GCH 2018

Eurographics Workshop on Graphics and Cultural Heritage (GCH) <16, 2018, Vienna, Austria>

3D scanning and 3D printing are two rapidly evolving domains, both generating results with a huge and growing spectrumof applications. Especially in Cultural Heritage, a massive and increasing amount of objects awaits digitization for variouspurposes, one of them being replication. Yet, current approaches to optical 3D digitization are semi-automatic at best andrequire great user effort whenever high quality is desired. With our solution we provide the missing link between both domains,and present a fully automatic 3D object replicator which does not require user interaction. The system consists of ourphotogrammetric 3D scanner CultArc3D_mini that captures an optimal image set for 3D geometry and texture reconstructionand even optical material properties of objects in only minutes, a conveyor system for automatic object feed-in and -out,a 3D printer, and our sensor-based process flow software that handles every single process step of the complex sequencefrom image acquisition, sensor-based object transportation, 3D reconstruction involving different kinds of calibrations, to3D printing of the resulting virtual replica immediately after 3D reconstruction. Typically, one-button machines require theuser to start the process by interacting over a user interface. Since positioning and pickup of objects is automatically registered,the only thing left for the user to do is placing an object at the entry and retrieving it from the exit after scanning. Shortly after,the 3D replica can be picked up from the 3D printer. Technically, we created a zero-button 3D replicator that provides highthroughput digitization in 3D, requiring only minutes per object, and it is publicly showcased in action at 3IT Berlin.

Show publication details
Volk, Rebekka; Luu, Thu Huong; Mueller-Roemer, Johannes; Sevilmis, Neyir; Schultmann, Frank

Deconstruction Project Planning of Existing Buildings Based on Automated Acquisition and Reconstruction of Building Information

2018

Automation in Construction

During their lifecycles, buildings are changed and adapted to the requirements of generations of users, residents and proprietaries over several decades. At the end of their life time, buildings undergo either retrofit or deconstruction (and replacement) processes. And, modifications and deviations of the original building structure, equipment and fittings as well as the deterioration and contamination of buildings are often not well documented or only available in an outdated and unstructured way. Thus, in many existing buildings, incomplete, obsolete or fragmented building information is predominating and hampering retrofit and deconstruction project planning. To plan change or deconstruction measures in existing buildings, buildings are audited manually or with stationary laser scans which requires great effort of skilled staff and expensive equipment. Furthermore, current building information models or deconstruction planning systems are often not able to deal with incomplete building information as it occurs in existing buildings. We develop a combined system named ResourceApp of a hardware sensor with software modules for building information acquisition, 3D reconstruction, object detection, building inventory generation and optimized project planning. The mobile and wearable system enables planner, experts or decision makers to inspect a building and at the same time record, analyze, reconstruct and store the building digitally. For this purpose, a Kinect sensor acquires point clouds and developed algorithms analyze them in real-time to detect construction elements. From this information, a 3D building model and building inventory is automatically derived. Then, the generated building reconstruction information is used for optimized project planning with a solution algorithm of the multi-mode resource-constrained project scheduling problem (MRCPSP) at hand. In contrast to existing approaches, the system allows mobile building recording during building walkthrough, real-time reconstruction and object detection. And, based on the automatically captured and processed building conditions by sensor data, the system performs an integrated project planning of the building deconstruction with available resources and the required decontamination and deconstruction activities. Furthermore, it optimizes time and cost considering secondary raw material recovery, usage of renewable resources, staff qualification, onsite logistics, material storage and recycling options. Results from field tests on acquisition, reconstruction and deconstruction planning are presented and discussed in an extensive non-residential case study. The case study shows that the building inventory masses are quite well approximated and project planning works well based on the chosen methods. Nevertheless, future testing and parameter adjustment for the automated data processing is needed and will further improve the systems' quality, effectiveness and accuracy. Future research and application areas are seen in the quantification and analysis of the effects of missing data, the integration of material classification and sampling sensors into the system, the system connection to Building Information Modelling (BIM) software via a respective interface and the transfer and extension to retrofit project planning.

Show publication details

Deep and Multi-algorithmic Gender Classification of Single Fingerprint Minutiae

2018

FUSION 2018

International Conference on Information Fusion (FUSION) <21, 2018, Cambridge, UK>

Accurate fingerprint gender estimation can positively affect several applications, since fingerprints are one of the most widely deployed biometrics. For example, gender classification in criminal investigations may significantly minimize the list of potential subjects. Previous work mainly offered solutions for the task of gender classification based on complete fingerprints. However, partial fingerprint captures are frequently occurring in many applications, including forensics and the fast growing field of consumer electronics. Moreover, partial fingerprints are not well-defined. Therefore, this work improves the gender decision performance on a well-defined partition of the fingerprint. It enhances gender estimation on the level of a single minutia. Working on this level, we propose three main contributions that were evaluated on a publicly available database. First, a convolutional neural network model is offered that outperformed baseline solutions based on hand crafted features. Second, several multi-algorithmic fusion approaches were tested by combining the outputs of different gender estimators that help further increase the classification accuracy. Third, we propose including minutia detection reliability in the fusion process, which leads to enhancing the total gender decision performance. The achieved gender classification performance of a single minutia is comparable to the accuracy that previous work reported on a quarter of aligned fingerprints including more than 25 minutiae.

  • 978-0-9964527-6-2
Show publication details
Damer, Naser; Wainakh, Yaza; Henniger, Olaf; Croll, Christian; Berthe, Benoit; Braun, Andreas; Kuijper, Arjan

Deep Learning-based Face Recognition and the Robustness to Perspective Distortion

2018

24th International Conference on Pattern Recognition. Proceedings

International Conference on Pattern Recognition (ICPR) <24, 2018, Beijing, China>

Face recognition technology is spreading into a wide range of applications. This is mainly driven by social acceptance and the performance boost achieved by the deep learningbased solutions in the recent years. Perspective distortion is an understudied distortion in face recognition that causes converging verticals when imaging 3D objects depending on the distance to the object. The effect of this distortion on face recognition was previously studied for algorithms based on hand-crafted features with a clear negative effect on verification performance. Possible solutions were proposed by compensating the distortion effect on the face image level, which requires knowing the camera settings and capturing a high quality image. This work investigates the effect of perspective distortion on the performance of a deep learning-based face recognition solution. It also provides a device parameter-independent solution to decrease this effect by creating more perspective-robust face representations. This was achieved by training the deep learning model on perspective-diverse data, without increasing the size of the training data. Experiments performed on the deep model in hand and a specifically collected database concluded that the perspective distortion effects face verification performance if not considered in the training process, and that this can be improved by our proposal of creating robust face representations by properly selecting the training data.

  • 978-1-5386-3787-6
Show publication details
Gutjahr, Christian; Kuijper, Arjan [1. Gutachten]; Rus, Silvia [2. Gutachten]

Designing Self-Aware Textiles

2018

Darmstadt, TU, Master Thesis, 2018

We are surrounded by textiles in our everyday live. Making them capable of local monitoring and computing is already a growing field of research in the area of Smart Home and Ambient Assisted Living. Equipping usual furniture with sensors and simple computational elements can provide useful information about the user and help with identifying emergency events, for instance fall recognition. This thesis investigates an approach to apply those ideas to textile materials worn by users by embedding inertial measurement unit sensors in a non-intrusive manner. In our approach, a simulation framework is used to ensure the highest possible accuracy while keeping the amount of sensors needed as low as possible. For this, a simulated sensor grid across the whole jacket was evaluated. Later, a prototype which uses the deformation of the jacket to provide valuable information about the current state of the jacket will be introduced. The presented use case to help find the jacket is just one idea on how to use the information gained by the sensor network.

Show publication details
Saeedan, Faraz; Weber, Nicolas; Goesele, Michael; Roth, Stefan

Detail-Preserving Pooling in Deep Networks

2018

2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) <2018, Salt Lake City, Utah, USA>

Most convolutional neural networks use some method for gradually downscaling the size of the hidden layers. This is commonly referred to as pooling, and is applied to reduce the number of parameters, improve invariance to certain distortions, and increase the receptive field size. Since pooling by nature is a lossy process, it is crucial that each such layer maintains the portion of the activations that is most important for the network’s discriminability. Yet, simple maximization or averaging over blocks, max or average pooling, or plain downsampling in the form of strided convolutions are the standard. In this paper, we aim to leverage recent results on image downscaling for the purposes of deep learning. Inspired by the human visual system, which focuses on local spatial changes, we propose detailpreserving pooling (DPP), an adaptive pooling method that magnifies spatial changes and preserves important structural detail. Importantly, its parameters can be learned jointly with the rest of the network. We analyze some of its theoretical properties and show its empirical benefits on several datasets and networks, where DPP consistently outperforms previous pooling approaches.

Show publication details
Lan, Zirui; Sourina, Olga; Wang, Lipo; Scherer, Reinhold; Müller-Putz, Gernot

Domain Adaptation Techniques for EEG-based Emotion Recognition: A Comparative Study onTwo Public Datasets

2018

IEEE Transactions on Cognitive and Developmental Systems

Affective brain-computer interface (aBCI) introduces personal affective factors to human-computer interaction. The state-of-the-art aBCI tailors its classifier to each individual user to achieve accurate emotion classification. A subject-independent classifier that is trained on pooled data from multiple subjects generally leads to inferior accuracy, due to the fact that encephalogram (EEG) patterns vary from subject to subject. Transfer learning or domain adaptation techniques have been leveraged to tackle this problem. Existing studies have reported successful applications of domain adaptation techniques on SEED dataset. However, little is known about the effectiveness of the domain adaptation techniques on other affective datasets or in a cross-dataset application. In this paper, we focus on a comparative study on several state-of-the-art domain adaptation techniques on two datasets: DEAP and SEED. We demonstrate that domain adaptation techniques can improve the classification accuracy on both datasets, but not so effective on DEAP as on SEED. Then, we explore the efficacy of domain adaptation in a cross-dataset setting when the data are collected under different environments using different devices and experimental protocols. Here, we propose to apply domain adaptation to reduce the intersubject variance as well as technical discrepancies between datasets, and then train a subject-independent classifier on one dataset and test on the other. Experiment results show that using domain adaptation technique in a transductive adaptation setting can improve the accuracy significantly by 7.25% – 13.40% compared to the baseline accuracy where no domain adaptation technique is used.

Show publication details
Krispel, Ulrich; Settgast, Volker; Fellner, Dieter W.

DynaMo - Dynamic 3D Models for the Web

2018

Proceedings Web3D 2018

International Conference on 3D Web Technology (WEB3D) <23, 2018, Poznań, Poland>

Animations aid greatly to a presentation of a sophisticated manmade object. With additional interactivity, a user can explore such an object to gain even better understanding. The authoring of such dynamic models is often very resource demanding, as the animations and logic of interaction has to be expressed in an authoring program; often using a programming language. Furthermore, the 3D model and its dynamic capabilites are tightly coupled, which makes it costly to integrate changes of the underlying 3D model - e.g. if a machine part changes. Contribution and Benefit. We present a novel scheme to define varying states of a 3D model in a decoupled, declarative manner using pattern matching. Furthermore, we demonstrate its capabilites for the web with an open source JavaScript implementation that operates on an x3dom model.

Show publication details

Efficient Global Registration for Nominal/Actual Comparisons

2018

VMV 2018

International Symposium on Vision, Modeling and Visualization (VMV) <23, 2018, Stuttgart, Germany>

We investigate global registration methods for Nominal/Actual comparisons, using precise, high-resolution 3D scans. First we summarize existing approaches and requirements for this field of application. We then demonstrate that a basic RANSAC strategy, along with a slightly modified version of basic building blocks, can lead to a high global registration performance at moderate registration times. Specifically, we introduce a simple feedback loop that exploits the fast convergence of the ICP algorithm to efficiently speed up the search for a valid global alignment. Using the example of 3D printed parts and range images acquired by two different high-precision 3D scanners for quality control, we show that our method can be efficiently used for Nominal/Actual comparison. For this scenario, the proposed algorithm significantly outperforms the current state of the art, with regards to registration time and success rate.

  • 978-3-03868-072-7
Show publication details

Efficient Pose Selection for Interactive Camera Calibration

2018

Proceedings of the 2018 IEEE International Symposium on Mixed and Augmented Reality

IEEE International Symposium on Mixed and Augmented Reality (ISMAR) <17, 2018, Munich, Germany>

The choice of poses for camera calibration with planar patterns is only rarely considered — yet the calibration precision heavily depends on it. This work presents a pose selection method that finds a compact and robust set of calibration poses and is suitable for interactive calibration. Consequently, singular poses that would lead to an unreliable solution are avoided explicitly, while poses reducing the uncertainty of the calibration are favoured. For this, we use uncertainty propagation. Our method takes advantage of a self-identifying calibration pattern to track the camera pose in real-time. This allows to iteratively guide the user to the target poses, until the desired quality level is reached. Therefore, only a sparse set of key-frames is needed for calibration. The method is evaluated on separate training and testing sets, as well as on synthetic data. Our approach performs better than comparable solutions while requiring 30% less calibration frames.

  • 978-1-5386-7459-8
Show publication details
Syed, Samar; Kuijper, Arjan [1. Gutachten]; Stein, Christian [2. Gutachten]; Sturm, Timo [3. Gutachten]

Ein generalisiertes Datenschema zur Einbettung visueller Meta-Informationen in Mixed-Reality-Anwendungen am Beispiel von Reparaturleitfäden

2018

Darmstadt, TU, Bachelor Thesis, 2018

Aufgabe dieser Arbeit war die Evaluation beziehungsweise Erweiterung eines vorgegebenen, generalisierten Datenschemas zur Definition von Reparaturleitfäden, in Hinblick auf seine Verwendbarkeit zur nutzerfreundlichen Einbettung visueller Meta-Informationen innerhalb von Mixed-Reality-Anwendungen. Der Fokus lag hierbei auf einer optimierten und automatischen Anordnung der, in Form von 2D-Annotationen gegebenen, textuellen Metainformationen, in Relation zu den verknüpften Mixed-Reality-3D-Inhalten.Neben der Evaluation des Datenschemas zur Bereitstellung der Inhalte und einer programmatischen Umsetzung ausgewählter Platzierungsalgorithmen wurde die Nutzerfreundlichkeit des Ergebnisses anhand einer Studie zur Bedienbarkeit gesichert.

Show publication details

Eliminating the Ground Reference for Wireless Electric Field Sensing

2018

Ambient Intelligence

European Conference on Ambient Intelligence (AmI) <14, 2018, Larnaca, Cyprus>

Capacitive systems are getting more and more attention these days. But many systems today like smart-phone screens, laptops, and non-mechanical buttons use capacitive techniques to measure events within several centimeters of distance. The reason that battery-powered devices don’t have high measurement ranges lies in the principle of capacitive measurement itself - the electrical ground is an inherent part of the measurement. In this paper, we present a method for passive and wireless capacitive systems to eliminate the reference to ground. This bears a couple of advantages for mobile, battery-powered capacitive sensor designs in the field of ambient intelligence. We compare the detection range of normal passive capacitive systems with our new approach. The results show that our improvements result in a higher detection range and higher power efficiency.

  • 978-3-030-03061-2
Show publication details
Frank, Sebastian; Kuijper, Arjan

Enabling Driver Feet Gestures Using Capacitive Proximity Sensing

2018

2018 International Conference on Intelligent Environments

International Conference on Intelligent Environments (IE) <14, 2018, Roma, Italy>

Due to driver assistance systems and the trend of industry towards automated driving, the hands and feet of the driver in a vehicle require less intervention, becoming even idle. Recent gesture recognition focuses on hand interaction. This paper provides feet gesture interaction. Many gesture recognition systems rely on computing intensive video systems, causing privacy concerns. Furthermore, these systems require a line of sight and therefore a visible interior design integration. Our system proves that invisibly integrated capacitive proximity sensors can do this as well. They do not cause privacy issues and can be integrated under non-conductive materials. Therefore, there is no impact on the design of the visible interior. Our proposed solution distinguishes between four feet gestures. There is no limitation to feet movement. Further, we prove the functionality of the system in an evaluation with six participants, using a prototypical mock-up of a vehicle legroom.

  • 978-1-5386-6844-3
Show publication details
Enimidisetty, Krishna Chaithanya; Kuijper, Arjan [1. Gutachten]; Ben Hmida, Helmi [2. Gutachten]

Enhacing the Interoperability Aspect of Formal and Informal Healthcare Systems

2018

Darmstadt, TU, Master Thesis, 2018

Increasing number of mobile devices (Personal Health Devices) are used at homes for first hand knowledge of vital signs like Blood pressure, glucose level, weight, temperature. The need to access this data directly by care givers (clinicians, doctors, hospitals) is becoming a norm. Personal Health devices (PHDs) can now be purchased at prices affordable to vast majority of people and ability to use them without any hindrance is reducing due to increase in focus on simple and intuitive user interface of these devices. It is estimated that a 300 million people will enter old age by 2025 [citation needed] and the need for affordable health care for patients and care givers at their fingertips will be the need of the hour. By collaborating with other ecosystems like smart home the number of use cases increases and will provide more incentive for general public to buy into the ecosystem. In this thesis functionality of the existing application, Smart Doktor enhanced by building a formal interoperable approach adhering to HL7 standard messaging and transport mechanisms. We extend the Smart Doktor application with two new vital signs namely ECG and PPG along with Smart devices namely, Smart Movement sensor and Smart Bed which provides a new insights into the patient under treatment and provides care givers ability to enhance the care they provide. An open source real dataset is used for implementing these devices using open source integration software tool “HL7Soup“.

Show publication details
Leissner, Johanna; Fuhrmann, Constanze

Europäisches Kulturerbejahr: Forschen und bewahren für die Zukunft

2018

ICOM Deutschland

Im Dezember 2017 wurde das Europäische Jahr des Kulturerbes eröffnet und imJanuar 2018 fand die deutsche Auftaktveranstaltung statt. Das Kulturerbe steht damitprominent im Mittelpunkt – eine Chance für mehr Forschung und Nachhaltigkeitzum Erhalt des kulturellen Erbes in Zeiten des Klimawandels.

Show publication details
Kolf, Jan Niklas; Wilmsdorff, Julian von [1. Gutachten]; Kuijper, Arjan [2. Gutachten]

Evaluation of Activity of Daily Life Recognition based on Electric Field Tokens

2018

Darmstadt, TU, Bachelor Thesis, 2018

Electric field sensors are used in a variety of ways to recognize different human actions and behaviors, for example, fall detection or classification of movements. However, very little is known about the number of sensors that are needed to achieve an acceptable recognition rate. Most systems just use as many sensors as possible to achieve confusion matrices with high true positive and true negative rates. In this thesis, the relation of recognition rates and the size of a system composed of electric field sensors shall be further investigated. For this purpose, several setups to recognize different human activities will be created and evaluated, each with a varying number of sensor tokens.

Show publication details
Breitfelder, Simon; Kuijper, Arjan [1. Gutachten]; Ritz, Martin [2. Gutachten]

Example-based Synthesis of Seamless Texture Variations and Application to the Acquisition of Optical Material-Properties

2018

Darmstadt, TU, Master Thesis, 2018

This work extends an existing workflow for acquisition, synthesis and rendering of Approximate Bi-directional Texturing Functions (ABTF), which represent a lower-dimensional alternative to Bi-directional Texturing Functions (BTF). In the first steps, image corrections and registration are presented to optimize the current setup. Texture synthesis needs to consider the surface geometry and reflection parameters to generate consistent images for all illumination directions. Furthermore, a method for generating seamless texture tiles and randomly spread them on a target surface to create renderings with less noticeable repetitions in texture is discussed in detail. To handle the enormous memory consumption of multiple ABTF datasets a fitting scheme is proposed specifically designed to reconstruct the most prominent effects captured in the ABTF model.

Show publication details
Müller, Maximilian; Kuijper, Arjan [Prüfer]; Bernard, Jürgen [Betreuer]

Explorative Analyse von multimodalen Patienten-Daten zur Erstellung und zum visuellen Vergleich von mehreren Patientenkohorten

2018

Darmstadt, TU, Master Thesis, 2018

Ärzte müssen oft entscheiden, welche Behandlung sie bei Patienten anwenden. Aufgrund der Vielzahl von Variablen und zahlreicher bereits vorhandenen Patientendaten ist die Suche nach Korrelationen und Zusammenhängen schwierig. Daher wurde ein Ansatz entwickelt, welcher Patientendaten so visualisiert, dass Ärzte herausfinden können, welche Behandlung als nächstes am erfolgversprechendsten ist oder welche Resultate nach einer bestimmten Behandlung zu erwarten sind. Der Ansatz visualisiert Event-Sequenzen und multimodale Attribute von Patienten gleichzeitig, um sie Suche nach Relationen zwischen ihnen zu ermöglichen. Ebenfalls können Nutzer Korrelationen zwischen Event-Sequenzen suchen, sowie Patienten-Kohorten beliebiger multimodaler Attribute erstellen und diese, sowie Teil- und Schnittmengen miteinander vergleichen. Dabei wurde der gesamte Ansatz so generisch gehalten, dass nicht nur Patientendaten, sondern beliebige temporale und multimodale Daten visualisiert werden können.

Show publication details

Exploring Accelerometer-based Step Detection by using a Wheeled Walking Frame

2018

iWOAR 2018

International Workshop on Sensor-based Activity Recognition (iWOAR) <5, 2018, Rostock, Germany>

ACM International Conference Proceedings Series

Step detection with accelerometers is a very common feature that smart wearables already include. However, when using a wheeled walking frame / rollator, current algorithms may be of limited use, since a different type of motion is being excreted. In this paper, we uncover these limitations of current wearables by a pilot study. Furthermore, we investigated an accelerometer-based step detection for using a wheeled walking frame, when mounting an accelerometer to the frame and at the user’s wrist. Our findings include knowledge on signal propagation of each axis, knowledge on the required sensor quality and knowledge on the impact of different surfaces and floor types. In conclusion, we outline a new step detection algorithm based on accelerometer input data. Our algorithm can significantly empower future off-the-shelf wearables with the capability to sufficiently detect steps with elderly people using a wheeled walking frame. This can help to evaluate the state of health with regard to the human behavior and motor system and even to determine the progress of certain diseases.

  • 978-1-4503-6487-4
Show publication details
Kügler, David; Distergoft, Alexander; Kuijper, Arjan; Mukhopadhyay, Anirban

Exploring Adversarial Examples Patterns of One-Pixel Attacks

2018

Understanding and Interpreting Machine Learning in Medical Image Computing Applications

International Workshop on Machine Learning in Clinical Neuroimaging (MLCN) <1, 2018, Granada, Spain>

Lecture Notes in Computer Science (LNCS)
11038

Failure cases of black-box deep learning, e.g. adversarial examples, might have severe consequences in healthcare. Yet such failures are mostly studied in the context of real-world images with calibrated attacks. To demystify the adversarial examples, rigorous studies need to be designed. Unfortunately, complexity of the medical images hinders such study design directly from the medical images. We hypothesize that adversarial examples might result from the incorrect mapping of image space to the low dimensional generation manifold by deep networks. To test the hypothesis, we simplify a complex medical problem namely pose estimation of surgical tools into its barest form. An analytical decision boundary and exhaustive search of the one-pixel attack across multiple image dimensions let us localize the regions of frequent successful one-pixel attacks at the image space.

  • 978-3-030-02627-1
Show publication details
End, Felix; Stork, André [1. Gutachten]; Grasser, Tim [2. Gutachten]

Exponential Integrators for Deformable Objects using Finite Elements

2018

Darmstadt, TU, Master Thesis, 2018

Simulating deformable objects is often done with implicit integrator since they can robustly simulate large time steps. However, these integrators introduce significant artificial damping into the simulation. Recently, exponential integrators have attracted attention as an alternative to implicit integrators. They have shown to the potential to achieve large time steps without introducing artificial damping. In this thesis we implement and compare five promising exponential integrator on deformable object simulation tasks. They are the non-linear exponential time integrator (NETI), co-rotational exponential time integrator (CETI), exponential Rosenbrock-Euler (ERE) and two exponential propagation iterative methods of Runge-Kutta type (EPIRK). Additionally, we also propose a new way to compute the exponential-like functions inside ERE and EPIRK integrators. Four of the exponential integrators we present have been used before for deformable object simulation but their performance has not been compared on the same task. In our experiment we compare how well the exponential integrators preserve the energy of the system, how accurate they are, their runtime requirements, how damping effects the integrator and how important internal variables scale with the task. We show that the exponential integrators preserve energy much better than the popular semi-implicit Euler integrator. Of the exponential integrators the ERE integrator is the most stable on our experiments and performs well in terms of accuracy. However, ERE is not as stable as the semi-implicit Euler integrator. Our new computation of the exponential functions in ERE and EPIRK slightly improves their stability but also slightly increases runtime requirements.

Show publication details
Demir, Ebru; Altenbernd, Peter [Referent]; Schütte, Alois [Korreferent]; Ritz, Martin [Betreuer]

Extension of the ABTF Material Acquisition and Rendering Process to CultArc3D Image Data

2018

Darmstadt, Hochschule, Bachelor Thesis, 2018

The importance of photorealistic 3D rendering of different materials is increasing, as there are various domains of application, such as the textile and the 3D games industry. In order to be able to do real-time rendering involving a physical material, a method for its acquisition and realistic rendering on 3D geometry is required. So far a single-camera system called ABTF Scanner is already able to acquire flat materials that are anisotropic (appearance dependent on rotation around surface normal) using a turntable, which makes it possible to map the acquired material onto an arbitrary 3D geometry during real-time rendering. Another scanner system consisting of multiple cameras called CultArc3D can also be used for this purpose. Due to its structure, allowing for lighting over a hemisphere by turning an arc equipped with light sources, there is no need for a turntable to acquire materials, as opposed to the single-camera system that achieves hemispheric illumination by combining a fixed quarter light arc with a rotary. In order to make images acquired by CultArc3D usable for real-time rendering, this thesis extends the software implemented for the ABTF Scanner. The extension was done in a way that makes the software multi functional in that it is now able to do real-time rendering for materials acquired by either of the above mentioned scanner system. For the first time images taken by CultArc3D can be used for renderings of ABTF material samples that capture material behavior for a comparable set of virtual light directions. Additionally a new shader (computer program used for 3D rendering) is implemented to provide real-time rendering with respect to the different data structure imposed by the concept of CultArc3D. The experimental evaluation shows that real-time rendering with the images acquired by CultArc3D can lead to better results compared to images taken by the ABTF Scanner, because the back-rotation of images, introduced by the rotary in the ABTF Scanner setup, is not required by CultArc3D. Thus, a number of calibrations and alignment steps that possibly introduce visual artifacts if not performed correctly, can be avoided. As a result CultArc3D can now be used for ABTF real-time rendering in addition to its capability of acquiring geometry, texture and a number of different optical material models. The software can be extended for different viewing perspectives during rendering in future work, due a hemispherical distribution of camera perspectives around the object.

Show publication details
Berthe, Benoit; Croll, Christian; Henniger, Olaf

Face Verification Robustness & Camera-Subject Distance

2018

Keesing Journal of Documents & Identity

After the international standardisation working group responsible for machine-readable travel documents (ISO/IEC JTC1/SC17/WG3) raised the question whether magnification distortion of facial images affected the accuracy of facial recognition, a study was conducted to determine to which degree the algorithms used for face comparison are affected. This study was described in a previous issue of Keesing's Journal of Documents and Identity.[1] It involved photographing several hundreds of enrolees at various distances (0.5 m to 3 m) with an automatic bench. One of the working group's findings discussed in this article is that magnification distortion of facial images due to pictures taken too close up does negatively affect facial recognition. However, for distances over 0.5 m the effect is very limited.

Show publication details
Cornel, Miriam; Kohlhammer, Jörn [1. Gutachten]; Gonzalez Parra, Garoe [2. Gutachten]

Facilitating Pilots’ Visual Assessment and Awareness of NOTAMs through a User-Centered Design Approach

2018

Darmstadt, TU, Master Thesis, 2018

The assessment of NOTAMs in flight preparation represents a big cognitive effort for pilots, due to their amount, the lack of standardization, prioritization and interaction mechanisms, and an often insufficient visual representation. This can result in important information inside the NOTAMs to be overlooked, and subsequently in the occurrence of hazardous situations. While supporting mechanisms are already being developed, the challenge of prioritizing NOTAMs is still not focused. By implementing a prototype of a system for the prioritization of NOTAMs and the adequate visual presentation thereof, this work aims at reducing the mental workload of pilots during flight preparation. As the underlying prioritization mechanism, Google’s PageRank algorithm[1] was used and transferred to the application field of aviation. Methods of user-centered design were utilized to implement a mechanism for prioritizing NOTAMs with the help of the PageRank algorithm in a way that supports pilots in their tasks and as a result, a prototype that presents the outcome by visual means was developed. Evaluating the system indicated that the chosen prioritization mechanism is a helpful addition to the ongoing efforts of making the interaction with NOTAMs easier for pilots.

Show publication details
Zhou, Wei; Ma, Caiwen; Liao, Shenghui; Shi, Jinjing; Yao, Tong; Chang, Peng; Kuijper, Arjan

Feature Fusion Information Statistics for Feature Matching in Cluttered Scenes

2018

Computers & Graphics

Object recognizing in cluttered scenes remains a largely unsolved problem, especially when applying feature matching to cluttered scenes there are many feature mismatches between the scenes and models. We propose our Feature Fusion Information Statistics (FFIS) as the calculation framework for extracting salient information from a Local Surface Patch (LSP) by a Local Reference Frame (LRF). Our LRF is defined on each LSP by projecting the scatter matrix’s eigenvectors to a plane which is perpendicular to the normal of the LSP. Based on this, our FFIS descriptor of each LSP is calculated, for which we use the combined distribution of mesh and point information in a local domain. Finally, we evaluate the speed, robustness and descriptiveness of our FFIS with the state-of-the-art methods on several public benchmarks. Our experiments show that our FFIS is fast and obtains a more reliable matching rate than other approaches in cluttered situations.

Show publication details

Fingerprint and Iris Multi-biometric Data Indexing and Retrieval

2018

FUSION 2018

International Conference on Information Fusion (FUSION) <21, 2018, Cambridge, UK>

Indexing of multi-biometric data is required to facilitatefast search in large-scale biometric systems. Previous worksaddressing this issue in multi-biometric databases focused onmulti-instance indexing, mainly iris data. Few works addressedthe indexing in multi-modal databases, with basic candidate listfusion solutions limited to joining face and fingerprint data. Irisand fingerprint are widely used in large-scale biometric systemswhere fast retrieval is a significant issue. This work proposes jointmulti-biometric retrieval solution based on fingerprint and irisdata. This solution is evaluated under eight different candidatelist fusion approaches with variable complexity on a databaseof 10,000 reference and probe records of irises and fingerprints.Our proposed multi-biometric retrieval of fingerprint and irisdata resulted in a reduction of the miss rate (1- hit rate) at 0.1%penetration rate by 93% compared to fingerprint indexing and88% compared to iris indexing.

Show publication details
Fu, Biying; Kirchbuchner, Florian; Kuijper, Arjan; Braun, Andreas; Gangatharan, Dinesh Vaithyalingam

Fitness Activity Recognition on Smartphones Using Doppler Measurements

2018

Informatics

Quantified Self has seen an increased interest in recent years, with devices including smartwatches, smartphones, or other wearables that allow you to monitor your fitness level. This is often combined with mobile apps that use gamification aspects to motivate the user to perform fitness activities, or increase the amount of sports exercise. Thus far, most applications rely on accelerometers or gyroscopes that are integrated into the devices. They have to be worn on the body to track activities. In this work, we investigated the use of a speaker and a microphone that are integrated into a smartphone to track exercises performed close to it. We combined active sonar and Doppler signal analysis in the ultrasound spectrum that is not perceivable by humans. We wanted to measure the body weight exercises bicycles, toe touches, and squats, as these consist of challenging radial movements towards the measuring device. We have tested several classification methods, ranging from support vector machines to convolutional neural networks. We achieved an accuracy of 88% for bicycles, 97% for toe-touches and 91% for squats on our test set.

Show publication details
Sablatnig, R. [Program Co-Chair]; Wimmer, M. [Program Co-Chair]; Fellner, Dieter W. [Proceedings Production Ed.]

GCH 2018

2018

Eurographics Workshop on Graphics and Cultural Heritage (GCH) <16, 2018, Vienna, Austria>

The objective of the conference is to introduce and showcase new techniques and applications for supporting Cultural Heritage information ranging from data acquisition, analysis and synthesis, 3D documentation, and data management, to new forms of interactive presentation and 3D printing solutions. Within Visual Heritage 2018, Eurographics GCH continues to provide a premier scientific forum to exchange novel ideas and techniques in research, education and dissemination of Cultural Heritage information, to transfer them into practice, and identify future research and application opportunities.

  • 978-3-03868-057-4
Show publication details
Fauser, Johannes; Stenin, Igor; Kristin, Julia; Klenzner, Thomas; Schipper, Jörg; Fellner, Dieter W.; Mukhopadhyay, Anirban

Generalized Trajectory Planning for Nonlinear Interventions

2018

OR 2.0 Context-Aware Operating Theaters, Computer Assisted Robotic Endoscopy, Clinical Image-Based Procedures, and Skin Image Analysis

International Workshop on OR 2.0 Context-Aware Operating Theaters (OR 2.0) <1, 2018, Granada, Spain>

Lecture Notes in Computer Science (LNCS)
11041

Minimally invasive procedures with flexible instruments suchas endoscopes, needles or drilling units are becoming more and more common.Their automated insertion will be standard across several applicationsin operation rooms of the future. In such scenarios regular replanningfor feasible nonlinear trajectories is a mandatory step towardautomation. However, state of the art methods focus on isolated solutionsonly. In this paper we introduce a generalized motion planningformulation in SE(3), regarding both position and orientation, that issuitable for these approaches. To emphasize the generalization of this formulationwe evaluate the performance of proposed Bidirectional RapidlyexploringRandom Trees (Bi-RRT) on four different clinical applications:Drilling in temporal bone surgery, trajectory planning for cardiopulmonaryendoscopy, automatic needle insertion for spine biopsy and livertumor removal. Experiments show that for all four scenarios the formulationis suitable and feasible trajectories can be planned successfully.

  • 978-3-030-01200-7
Show publication details
Vashishtha, Anand; Kuijper, Arjan [1. Prüfer]; Mukhopadhyay, Anirban [2. Prüfer]

Generalized U-Net for Multi-Channel Image Segmentation

2018

Darmstadt, TU, Master Thesis, 2018

Image segmentation is a digital image processing technique which extracts important structural features of the image. Few of the practical applications of image segmentation are medical imaging, satellite imaging, remote sensing in the ecological domain and many others. U-Net[37] is a robust encoder-decoder neural network for pixel-wise segmentation of grayscale biomedical images. However, U-Net does not generalize for multi-channel image segmentation; a necessity in multiple applications, for example, Wireless Capsule Endoscopy (WCE) in medicine, and habitat interpretation in ecology. In this thesis, we propose a deep learning method based on U-Net to segment multi-channel images in both ecology and medical imaging. Biological diversity is decreasing at a rate of 100-1000 times pre-human rates [6][39], and to avoid species extinction; we need to understand factors influencing the occurrence of species. Fast, reliable computer-assisted tools can help to describe the habitat and thus to understand species habitat associations. This understanding is of utmost importance for more targeted species conservation efforts. Due to logistical challenges and time-consuming manual processing of field data, months up to years are often needed to progress from data collection to data interpretation. Image segmentation of vegetation dataset is one of the first tasks to interpret habitation preferences. Deep learning can be used to significantly shorten the time while keeping a similar level of accuracy. Here, we propose Habitat-Net: a novel Convolutional Neural Network (CNN) based method inspired from U-Net to segment multi-channel habitat images of rainforests. Compared to manual segmentation, Habitat-Net prediction is approximately 3K-150K times faster resulting in a significant increase of the processing time. Similarly, a typical wireless capsule endoscopy procedure generates tens of thousands of images, resulting in a manual diagnosis of small bowel diseases laborious and time-consuming. Recent automatic approaches focus on learning based methodologies for patch-level localization of abnormalities, resulting in the argument about pixel-level localization marginal. This, however, results in a gross under-estimation as the distinctive bleeding pattern of many small bowel diseases are sprinkled over and covers only a small surface area. We present CE-Net a novel generalization of U-Net [37] combining two regularization techniques, namely batch normalization and data augmentation for color image segmentation to automate pixel-level localization of the malign areas of WCE images. The advantages of CE-Net are demonstrated on a dataset containing 405 images across seven diseases, where CE-Net outperforms current state-of-the-art WCE image segmentation method in terms of Dice accuracy.

Show publication details
Mäs, Stephan; Henzen, Daniel; Bernard, Lars; Müller, Matthias; Jirka, Simon; Senner, Ivo

Generic Schema Descriptions for Comma-Separated Values Files of Environmental Data

2018

Geospatial Technologies for All

Conference on Geographic Information Science (AGILE) <21, 2018, Lund, Sweden>

Comma-Separated Values (CSV) files are commonly used to publish data about environmental phenomena and environmental sensor measurements. Due to its simplicity, this format has many advantages. However, at the same time there is no official standard for CSV and no possibility to specify schematic constraints or other metadata. As a result, CSV files come in many variations and often with no metadata that would support interpretation or further processing, analysis and visualization. In this paper, we propose a framework for the specification of schema descriptions for CSV files as they are used in the environmental sciences. It allows to constrain the structure and content of a CSV file and also to specify relations between files, for example when they are published in one data package. The framework is extensible, also to other spatial data formats such as GeoTiff. The schema descriptions are encoded in JSON or XML to be published in the Web as a supplement to the data. It comes as a lightweight solution that provides metadata required to publish OGC compliant services from CSV files. It helps to overcome the heterogeneities of different data providers when exchanging environmental measurement data on the Web.Keywords: tabular data, generic schema language, CSV, comma separated values, metadata.

Show publication details

GPU-based Polynomial Finite Element Matrix Assembly for Simplex Meshes

2018

Computer Graphics Forum

Pacific Conference on Computer Graphics and Applications (PG) <26, 2018, Hong Kong, China>

In this paper, we present a matrix assembly technique for arbitrary polynomial order finite element simulations on simplex meshes for graphics processing units (GPU). Compared to the current state of the art in GPU-based matrix assembly, we avoid the need for an intermediate sparse matrix and perform assembly directly into the final, GPU-optimized data structure. Thereby, we avoid the resulting 180% to 600% memory overhead, depending on polynomial order, and associated allocation time, while simplifying the assembly code and using a more compact mesh representation. We compare our method with existing algorithms and demonstrate significant speedups.

Show publication details
Lotter, Lars; Stork, André [1. Gutachten]; Mueller-Roemer, Johannes [2. Gutachten]

GPU-basierte Shallow Water Simulation

2018

Darmstadt, TU, Bachelor Thesis, 2018

Menschliche Eingriffe in natürliche Flussläufe sowie die steigende Urbanisierung verstärken das Risiko von Schäden durch Hochwasser insbesondere in bevölkerungsreichen Gebieten. Vergangene Ereignisse wie das Elbhochwasser von 2002, 2006, 2013 und im Sommer 2016 haben gezeigt, dass das Risiko für Mensch und Material durch Hochwasser in den letzten Jahren gestiegen ist. Zur Unterstützung in der Planung für den Katastrophenfall, sowie in der Vorhersage, werden numerische Simulationen genutzt, um den Wasserverlauf sowie das Abflussverhalten abzuschätzen. Um in der Vorhersage nutzbar zu sein, müssen diese Simulationen schneller als Echtzeit laufen. Da die zu simulierenden Gebiete oft mehrere Quadratkilometer abdecken, werden dazu in der Regel Approximationen wie die Shallow Water Equations, auch unter dem Namen Saint-Venant-Gleichungen bekannt, eingesetzt (siehe z.B. [1]). Dabei handelt es sich um eine Annäherung der Navier-Stokes-Gleichungen für allgemeine Fluiddynamik für Fälle in denen der Druck näherungs-weise hydrostatisch ist und das Fluidvolumen durch ein Höhenfeld beschrieben werden kann. Neben Approximationen ist es wichtig, die vorhandene Hardware optimal zu nutzen. Sei es durch spezialisierte Datenstrukturen (siehe z.B. [2]) oder durch Parallelisierung auf massiv parallelen Graphikprozessoren (GPUs), wie es Brodtkorb et al. [3] und Vacondio et al. [4] demonstriert haben. Ziel dieser Arbeit ist es, eine Auswahl an Verfahren bezüglich ihrer Eignung für die Berechnung auf GPUs zu treffen, diese Auswahl zu implementieren und in Bezug auf Geschwindigkeit und Präzision zu vergleichen. Dabei sind nur Verfahren, die auch Nass-Trocken-Fronten beschreiben können, in Betracht zu ziehen, da diese in der Überflutungssimulation notwendig sind. Literaturverzeichnis: [1] F. Aureli, A. Maranzoni, P. Mignosa und C. Ziveri, „A weighted surface-depth gradient method for the numerical integration of the 2D shallow water equations with topography,“ Advances in Water Resources, Bd. 31, Nr. 7, pp. 962-974, 2008. [2] L. Qiuhua und A. Borthwick, „Adaptive quadtree simulation of shallow flows with wet-dry fronts over,“ Computers & Fluids, Bd. 38, pp. 221-234, 2009. [3] A. Brodtkorb, M. Sætra und M. Altinakar, „Efficient shallow water simulations on GPUs: Implementation, visualization, verification, and validation,“ Computers & Fluids, Bd. 55, pp. 1-12, 2012. [4] R. Vacondio, A. Dal Palù und P. Mignosa, „GPU-enhanced Finite Volume Shallow Water solver for fast flood simulations,“ Environmental Modeling & Software, Bd. 57, pp. 60-75, 2014.

Show publication details
Zhou, Wei; Ma, Caiwen; Kuijper, Arjan

Hough-space-based Hypothesis Generation and Hypothesis Verification for 3D Object Recognition and 6D Pose Estimation

2018

Computers & Graphics

Hypothesis Generation (HG) and Hypothesis Verification (HV) play an important role in 3D objection recognition. However, performing 3D object recognition in cluttered scenes using HG and HV still re- mains a largely unsolved problem. High False Positive (FP) in HG and HV stages are witnessed due to clutter and occlusion, which will further affect the final accuracy of recognition. To address these prob- lems, we propose a novel Hough-space-based HG approach for extracting hypotheses. Differently from the existing methods, our approach is based on a Hough space which adopts a self-adapted measure to generate hypotheses. Based on this, a novel HV-based method is proposed to verify the hypotheses obtained from HG procedures. The proposed method is evaluated on four public benchmark datasets to verify its performance. Experiments show that our approach outperforms state-of-the-art methods, and obtains a higher recognition rate without sacrificing precision both at high FP rates and high occlusion rates.

Show publication details
Sah, Ashish Prasad; Kuijper, Arjan [Prüfer]; Fu, Biying [Betreuerin]

Human Activity Recognition Using Single Wire Electrode Based on Electric Potential Sensing

2018

Darmstadt, TU, Master Thesis, 2018

Our human body is the storehouse of electric potential. When there is a physiological response from a body, atoms are set in motion tends to generate an electric potential. These could be captured in the form of electric signals which shows visible variations as the body responds to certain motions. Observing the variation in these electric signals during movements is the direction that is followed in this thesis.Human body surface emits an electric signal which could be captured and could be studied upon for various purposes, such as electrocardiogram for observing electrical activity of the heart; electroencephalogram for the neurons firing in the brain; and the electroretinogram studying the electrical response of the cells in the retina. Similarly, these electrical potential generated by the human body can also be used to recognize activities using Electric Potential Sensing (EPS).EPS is a sensor based electric field measurement technique. It is a purely passive capacitive measurement technique, which requires extremely low power consumption for its operations. It passively monitors the user's activities all the time, thus not requiring any actions from the user to perform.The main aim of this thesis is to build a smart environment to recognize human activities thus leading to provide necessary support, especially to our elders to live an independent life. This system will be able to detect possible emergencies in case of fall.Hence, a smart environment is built with the help of a single wire connected to the sensor. Multiple experiments, in the various environmental setting for different scenarios, is conducted. This is being done to evaluate the performance of the sensor by testing it in all kinds of environmental settings and situations.Finally, after analyzing the potential of the system, SenseCare a real-time activity recognition system is developed to predict the activities as and when performed by the user. It is tuned for real-time performance and robustness by combining the state machine transition rules along with the classifier trained with the machine learning algorithm. Also, an alarm is raised by contacting the emergency contact of the user in case of an emergency.

Show publication details
Noll, Matthias; Noa-Rudolph, Werner; Wesarg, Stefan; Kraly, Michael; Stoffels, Ingo; Klode, Joachim; Spass, Cédric; Spass, Gerrit

ICG based Augmented-Reality-System for Sentinel Lymph Node Biopsy

2018

Eurographics Workshop on Visual Computing for Biology and Medicine

Eurographics Workshop on Visual Computing for Biology and Medicine (VCBM) <8, 2018, Granada, Spain>

In this paper we introduce a novel augmented-reality (AR) system for the sentinel lymph node (SLN) biopsy. The AR system consists of a cubic recording device with integrated stereo near-infrared (NIR) and stereo color cameras, an head mounted display (HMD) for visualizing the SLN information directly into the physicians view and a controlling software application. The labeling of the SLN is achieved using the fluorescent dye indocyanine green (ICG). The dye accumulates in the SLN where it is excited to fluorescence by applying infrared light. The fluorescence is recorded from two directions by the NIR stereo cameras using appropriate filters. Applying the known rigid camera geometry, an ICG depth map can be generated from the camera images, thus creating a live 3D representation of the SLN. The representation is then superimposed to the physicians field of view, by applying a series of coordinate system transformations, that are determined in four separate system calibration steps. To compensate for the head motion, the recording systems is continuously tracked by a single camera on the HMD using fiducial markers. Because the system does not require additional monitors, the physicians attention is kept solely on the operation site. This can potentially decrease the intervention time and render the procedure safer for the patient.

  • 978-3-03868-056-7
Show publication details
Prajapati, Ashok; Kuijper, Arjan [1. Gutachten]; Siegmund, Dirk [2. Gutachten]

Identifying Cuts and Holes in Fabrics

2018

Darmstadt, TU, Master Thesis, 2018

Quality assurance of fabrics is one of the basic and vital tasks in the textile industries. A human operated task can be error prone. Automatic visual inspection reduces a lot of time as well as the labour cost. Most of the approaches so far have been implemented over the flat spread textiles [35][20]. Main goal of this thesis is to detect and classify small or fine defects (cuts,holes,stains) as far as possible in in-homogeneous, voluminous shape fabrics. Similarly, The main focus is also on test time computation by minimizing the processing steps. To achieve this goal, deep learning (DL) and computer vision techniques are implemented which seem to be effective in the areas of image classification and object localization. This report will provide a detail overview about how an object detector algorithm can be used to detect defects and localize over the fabric. Most of the approach only classify the defects but not localize them as a part of the network itself. To obtain optimum classification accuracy and the computational cost, I train the RCNN(Faster)[31] model with the labelled defects(bounding box) in the images as an input so that the defect detection becomes real time. The feature maps are extracted from the last convolution layer of the Convolutional Neural Network (CNN) and then classified with the softmax layer (at the end of the fully-connected layer). The regression layer outputs the coordinates of the bounding box. Here, I use different CNN [22] architecture and the good result is obtained with the VGG16[37] pretrained model without using disparity map. The details about the methods and evaluation will be presented in subsequent methodology and evaluation respectively. The classification accuracy of 98.05% and 96.70% were obtained in the test set (at threshold 0.5) and in the validation set(at threshold 0.7) respectively with the proposed method.

Show publication details
Brakowski, Alexander; Kuijper, Arjan [1. Gutachten]; May, Thorsten [2. Gutachten]

Interactive Exploration of Personal Health Options for Prostate Cancer Patients

2018

Darmstadt, TU, Master Thesis, 2018

The impact of prostate cancer is significant on our society, since 1 in 8 men and their families will need to have to deal with the disease and its treatment at some point in their life. It is the second most common cancer among men in the world, with more than 1.1 million new cases a year world-wide. Mostly newly diagnosed prostate cancer patients are in the early-stage of the disease. Newly diagnosed patients could simply take a passive role in that whole treatment process, and let a doctor make all the decisions for them. Fortunately research shows, that most prostate cancer patients like to play an active role in the decision-making of their treatment, and like to know as much as possible about the disease. The problem is that prostate cancer is a highly heterogeneous disease with so many treatment options available, all with their own set of side effects, that making a decision is very hard. This problem motivated us to use the personal health data of similiar patients from PROGETHER, to help patient make a decision. We use that data to show users what most other patients did next. In this thesis we present a web-based application, that tries to solve that problem by allowing users to explore their treatment options in an interactive and user-friendly way by using the data of patients with similar treatment histories provided by PROGETHER. Because of the complexity of the data the application shows these data in a multi step process in form of a mixture of visualizations and easy understandable texts. Each step of that process handles a different part of the data, so that users can dive deeper into the data, the further that progress into the process. The application is carefully designed to match the needs of our target group, who are generally prostate cancer patients, over the age of 65, and not very tech-savvy. Special focus was put on the layout, colors and navigation of the application. We evaluated our approach by using a quantifiable user experience evaluation. The application was well accepted and user found it to be innovative, create, attractive and enjoyable, they understood it quite well and were able to solve their task fast and effortless.

Show publication details
Ferreira, Stephanie; Stork, André [1. Gutachten]; Weber, Daniel [2. Gutachten]

Interactive Plastic Deformation Simulation for Editing Surface and Tetrahedral Meshes

2018

Darmstadt, TU, Bachelor Thesis, 2018

In this thesis a physically based approach for editing surface and tetrahedral meshes is implementedand evaluated. Therefore an existing framework for the simulation of elasticity with the finite elementmethod and implicit time integration is enhanced by adding plasticity. The to be edited object mesh,given as surface or tetrahedral mesh, is deformed by coupling it to a coarser simulation mesh on whichthe simulation of elasto-plastic deformation is performed. As plastic deformation is not reversed, a newrest state is obtained by applying forces to the simulation mesh. The method is benchmarked with amedical model of a human consisting of sub-meshes depicting bones and muscles.

Show publication details
Chegini, Mohammad; Shao, Lin; Gregor, Robert; Lehmann, Dirk J.; Schreck, Tobias

Interactive Visual Exploration of Local Patterns in Large Scatterplot Spaces

2018

Computer Graphics Forum

Eurographics / IEEE VGTC Conference on Visualization (EuroVis) <20, 2018, Brno, Czech Republic>

Analysts often use visualisation techniques like a scatterplot matrix (SPLOM) to explore multivariate datasets. The scatterplots of a SPLOM can help to identify and compare two-dimensional global patterns. However, local patterns which might only exist within subsets of records are typically much harder to identify and may go unnoticed among larger sets of plots in a SPLOM. This paper explores the notion of local patterns and presents a novel approach to visually select, search for, and compare local patterns in a multivariate dataset. Model-based and shape-based pattern descriptors are used to automatically compare local regions in scatterplots to assist in the discovery of similar local patterns. Mechanisms are provided to assess the level of similarity between local patterns and to rank similar patterns e_ectively. Moreover, a relevance feedback module is used to suggest potentially relevant local patterns to the user. The approach has been implemented in an interactive tool and demonstrated with two real-world datasets and use cases. It supports the discovery of potentially useful information such as clusters, functional dependencies between variables, and statistical relationships in subsets of data records and dimensions.

Show publication details
Hartung, Daniel; Kuijper, Arjan [Prüfer]; Lücke-Tieke, Hendrik [Betreuer]

Interaktive Visualisierungen in flussdiagrammbasierten Programmierumgebungen

2018

Darmstadt, TU, Bachelor Thesis, 2018

Die Analyse und Verarbeitung großer Datenmengen stellt eine aktuelle Herausforderung dar, der sich sowohl Domainexperten als auch Laien stellen müssen. Eine für Laien und Experten etablierte Möglichkeit sind flussdiagrammbasierte Programmierumgebungen, mit denen sich eine Berechnungspipeline erstellen lässt. Diese Berechnungspipeline ist dabei in vorhandenen Lösungen azyklisch, was die Möglichkeiten für interaktive Visualisierungen einschränken kann. In dieser Arbeit soll erstmals ein prototypischer Ansatz entwickelt werden, mit dem diese konzeptionelle Beschränkung geschwächt wird. Dadurch ist es möglich die visuelle Interaktionsfähigkeit zu erhöhen. Insbesondere ein Brushing and Linking lässt sich so über das ,,physische'' Verbinden von Diagrammen realisieren. Eine qualitative Evaluation zeigt, dass diese Interaktionsmöglichkeiten einen sinnvollen Beitrag zur Datenanalyse bilden. Zudem erscheinen Zyklen für die Nutzer ein intuitives Werkzeug zu sein, um Brushing and Linking zu realisieren. In dieser Arbeit wird auf vorhandene Lösungen, die Ansätze und die Implementierung des Prototypens eingegangenen. Abschließend wird dieser qualitativ evaluiert und Limitationen des Ansatzes aufgezeigt.

Show publication details

Intervention Assessment Tool for Primary Tumors in the Liver

2018

Current Directions in Biomedical Engineering

After a liver tumor intervention the medical doctor has to compare both pre and postoperative CT acquisitions to ensure that all carcinogenic cells are destroyed. A correct assessment of the intervention is of vital importance, since it will reduce the probability of tumor recurrence. Some methods have been proposed to support the medical doctors during the assessment process, however, all of them focus on secondary tumors. In this paper a tool is presented that enables the outcome validation for both primary and secondary tumors. Therefore, a multiphase registration (preoperative arterial and portal phases) followed by a registration between the pre and postoperative CT images is carried out. The first registration is in charge of the primary tumors that are only visible in the arterial phase. The secondary tumors will be incorporated in the second registration step. Finally, the part of the tumor that was not covered by the necrosis is quantified and visualized. The method has been tested in 9 patients, with an average registration error of 1.41 mm.

Show publication details
Braun, Andreas; Zander-Walz, Sebastian; Majewski, Martin; Kuijper, Arjan

Investigating Large Curved Interaction Devices

2018

Personal and Ubiquitous Computing

Large interactive surfaces enable novel forms of interaction for their users, particularly in terms of collaborative interaction. During longer interactions, the ergonomic factors of interaction systems have to be taken into consideration. Using the full interaction space may require considerable motion of the arms and upper body over a prolonged period of time, potentially causing fatigue. In this work, we present Curved, a large-surface interaction device, whose shape is designed based on the natural movement of an outstretched arm. It is able to track one or two hands above or on its surface by using 32 capacitive proximity sensors. Supporting both touch and mid-air interaction can enable more versatile modes of use. We use image processing methods for tracking the user's hands and classify gestures based on their motion. Virtual reality is a potential use case for such interaction systems and was chosen for our demonstration application. We conducted a study with ten users to test the gesture tracking performance, as well as user experience and user preference for the adjustable system parameters.

Show publication details
Modenese, Luca; Montefiori, Erica; Wang, Anqi; Wesarg, Stefan; Viceconti, Marco; Mazzà, Claudia

Investigation of the Dependence of Joint Contact Forces on Musculotendon Parameters Using a Codified Workflow for Image-based Modelling

2018

Journal of Biomechanics

The generation of subject-specific musculoskeletal models of the lower limb has become a feasible task thanks to improvements in medical imaging technology and musculoskeletal modelling software. Nevertheless, clinical use of these models in paediatric applications is still limited for what concerns the estimation of muscle and joint contact forces. Aiming to improve the current state of the art, a methodology to generate highly personalized subject-specific musculoskeletal models of the lower limb based on magnetic resonance imaging (MRI) scans was codified as a step-by-step procedure and applied to data from eight juvenile individuals. The generated musculoskeletal models were used to simulate 107 gait trials using stereophotogrammetric and force platform data as input. To ensure completeness of the modelling procedure, muscles’ architecture needs to be estimated. Four methods to estimate muscles’ maximum isometric force and two methods to estimate musculotendon parameters (optimal fiber length and tendon slack length) were assessed and compared, in order to quantify their influence on the models’ output. Reported results represent the first comprehensive subject-specific model-based characterization of juvenile gait biomechanics, including profiles of joint kinematics and kinetics, muscle forces and joint contact forces. Our findings suggest that, when musculotendon parameters were linearly scaled from a reference model and the muscle force-length-velocity relationship was accounted for in the simulations, realistic knee contact forces could be estimated and these forces were not sensitive the method used to compute muscle maximum isometric force.

Show publication details
Manna, Subhadeep; Weber, Daniel [Advisor]; Stork, André [Supervising Professor]

Jacobi Method for Distributed Sparse Linear Systems on Multiple GPUs using CUDA

2018

Darmstadt, TU, Master Thesis, 2018

Modern graphics processing units (GPUs) having many-core architecture are capable of accelerating simulation applications based on Parallel Differential Equation (PDE) tremendously. PDE solvers are the backbone of physics-based simulations and are used in wide variety of applications such as Computational Fluid Dynamics (CFD), computer games, augmented and virtual environments. Modern high-performance computing (HPC) architectures are equipped with multiple GPUs. This pushes the development of parallel algorithms towards solving fine resolution problems which are often inherently large for a single GPU. A Jacobi iterative solver is a popular PDE solver for Poisson equation.This thesis focuses on the scalability of a single GPU Jacobi solver over multiple GPUs. Transferring of data between multiple GPUs in a node remains the primary bottleneck and thus prevents from developing an efficient solver. Redesigning a solver which works on a single GPU and scaling it over multiple GPUs is challenging. Such a task would need identifying and exploiting parallelism opportunities and minimizing the communication latency when transferring data from one GPU to another. Existing methods used for multi-GPU communication are analyzed and optimal approaches are suggested for exploiting the inter-GPU memory bandwidth. Different domain decomposition methods are also suggested and the implications on performance are evaluated. Time taken to solve a problem increases with the resolution of the problem on a single GPU. Dividing the problem into smaller parts and distributing them to multiple GPUs could result in faster computation times. The implications of distributing such a problem are evaluated by measuring computation times and communication latency. Thereby, suggesting ways to optimize the solver by increasing the parallelism.

Show publication details
Fellner, Dieter W. [Hrsg.]; Welling, Daniela [Red.]; Ackeren, Janine van [Red.]; Kröchert, Charlott [Red.]; Prasche, Svenja [Red.]; Roth, Anahit [Red.]; Egner, Julian [Gestaltung]; Gollnast, Anja [Gestaltung]; Jahnke, Anne [Gestaltung]; Kunkel, Andreas [Red.]

Jahresbericht 2017

2018

Das Fraunhofer IGD hat seine Forschungsaktivitäten vor Kurzem in vier Leitthemen gebündelt, welche die Basis seiner Arbeit bilden und verschiedene Themen abteilungsübergreifend miteinander verknüpfen. Eines dieser Leitthemen ist "Visual Computing as a Service - Die Plattform für angewandtes Visual Computing". Die Basis dieser universellen Plattform für Visual-Computing Lösungen ist gelegt und wird kontinuierlich erweitert. Dieser technologische Ansatz bildet die Grundlage für die weiteren Leitthemen. In der "Individuellen Gesundheit - Digitale Lösungen für das Gesundheitswesen" werden die Daten betrachtet, die in der personalisierten Medizin anfallen - mithilfe der Visual-Computing-Technologien des Instituts. Im Leitthema "Intelligente Stadt - Innovativ, digital und nachhaltig" ist die Fragestellung, wie man den Lebenszyklus urbaner Prozesse unterstützen kann. Und im Leitthema "Digitalisierte Arbeit - Der Mensch in der Industrie 4.0" geht es erster Linie um die Unterstützung des Menschen in der durch die Digitalisierung veränderten Produktion.

Show publication details
Tack, A.; Mukhopadhyay, Anirban; Zachow, S.

Knee Menisci Segmentation Using Convolutional Neural Networks: Data from the Osteoarthritis Initiative

2018

Osteoarthritis and Cartilage

Objective: To present a novel method for automated segmentation of knee menisci from MRIs. To evaluate quantitative meniscal biomarkers for osteoarthritis (OA) estimated thereof. Method: A segmentation method employing convolutional neural networks in combination with statistical shape models was developed. Accuracy was evaluated on 88 manual segmentations. Meniscal volume, tibial coverage, and meniscal extrusion were computed and tested for differences between groups of OA, joint space narrowing (JSN), and WOMAC pain. Correlation between computed meniscal extrusion and MRI Osteoarthritis Knee Score (MOAKS) experts' readings was evaluated for 600 subjects. Suitability of biomarkers for predicting incident radiographic OA from baseline to 24 months was tested on a group of 552 patients (184 incident OA, 386 controls) by performing conditional logistic regression. Results: Segmentation accuracy measured as dice similarity coefficient was 83.8% for medial menisci (MM) and 88.9% for lateral menisci (LM) at baseline, and 83.1% and 88.3% at 12-month follow-up. Medial tibial coverage was significantly lower for arthritic cases compared to non-arthritic ones. Medial meniscal extrusion was significantly higher for arthritic knees. A moderate correlation between automatically computed medial meniscal extrusion and experts' readings was found (r ¼ 0.44). Mean medial meniscal extrusion was significantly greater for incident OA cases compared to controls (1.16 ± 0.93 mm vs 0.83 ± 0.92 mm; P < 0.05). Conclusion: Especially for medial menisci an excellent segmentation accuracy was achieved. Our meniscal biomarkers were validated by comparison to experts' readings as well as analysis of differences w.r.t groups of OA, JSN, and WOMAC pain. It was confirmed that medial meniscal extrusion is a predictor for incident OA.

Show publication details
Leissner, Johanna; Fuhrmann, Constanze

Kulturerbe und Klimawandel - sind wir an einem Wendepunkt?

2018

Cartaditalia

Unser Kulturerbe lässt uns unsere Geschichte und unsere Zukunft verstehen. Deshalb wird 2018 das Europäische Kulturerbejahr ausgerufen, um seine vielschichtige Bedeutung für Europa und seine Bürger zu stärken und hervorzuheben. Doch dieses Kulturerbe ist mannigfaltigen Gefahren ausgesetzt und bedarf unseres Schutzes. Verfall, Umwelteinwirkungen, Vernachlässigung, bewaffnete Konflikte, Naturkatastrophen Abriss oder Umnutzung - die Ursachen für einen Verlust unwiederbringlicher Zeugnisse der Vergangenheit sind vielfältig (UNESCO, 2008a, b). Aber nun steht der Schutz unseres kulturellen Erbes an einem Wendepunkt. Eine noch nie dagewesene Gefahr von unvorstellbarem Ausmaß bedroht unsere Kulturgüter (IPCC 2007; UNESCO 2007; UNEP, UNESCO, Union of Concerned Scientists, 2016). Diese Gefahr ist der menschengemachte Klimawandel, der schon heute erhebliche Risiken für die Menschheit mit sich bringt. Im Dezember 2015 haben im Pariser Klimaabkommen 196 Staaten erstmals den menschengemachten Klimawandel anerkannt und beschlossen, die Treibhausgase zu reduzieren. Der Klimawandel bedroht Kulturerbe auf unterschiedliche Weise. Einerseits durch direkten Einfluss der extremen Wetter- und Umweltbedingungen (Cassar, 2005; English Heritage, 2008a). Auf der anderen Seite durch indirekte Folgen, die noch bedrohlicher ausfallen können: Entvölkerung ganzer Landstriche durch Klimaflüchtlinge, Destabilisierung sozialer Systeme, abnehmende finanzielle Ressourcen für Erhaltungsmaßnahmen oder das Ausbleiben von Besuchern und Touristen wegen extremer Wetterverhältnisse. Diese negativen Einflüsse des Klimawandels betreffen die Kulturgüter nicht nur selbst, indem sie ihre Integrität oder Existenz bedrohen. Sie haben auch weitreichende Konsequenzen für die sie umgebenden sozialen und ökonomischen Strukturen. Kulturgüter sind als Symbole und Identitätsanker wichtige Bezugspunkte menschlicher Kultur. Verlust oder Beschädigung dieser nicht erneuerbaren Ressourcen können gravierende Auswirkungen auf das auf sie Bezug nehmende Kollektiv haben. Kulturgüter und Denkmäler sind zudem bedeutende Einnahmequellen für lokale Ökonomien. Die (Kultur)Tourismusbranche ist besonders in ländlichen, strukturschwachen Regionen einer der wichtigsten Wirtschaftsfaktoren mit hoher Wertschöpfung und vielen Arbeitsplätzen. Ein Wegbrechen dieser Einnahmequelle wird verheerende wirtschaftliche Folgen mit sich bringen. Angesichts der gravierenden Auswirkungen, die der Klimawandel auf unser Kulturerbe hat, ist es verwunderlich, welch geringe Rolle diese Gefahren in Politik, Forschung und bei den für seinen Schutz verantwortlichen Institutionen noch immer spielen. Zwar wächst vereinzelt das Bewusstsein für die potenziellen Risiken, von flächendeckenden und vor allem koordinierten Aktivitäten und vorausschauenden Strategien wie die Folgen des Klimawandels abgemildert werden können, ist aber wenig bekannt. Zwar hat der Europarat (Council of Europe, 2017) im Juni 2017 eine "Europäische Kulturerbe Strategie für das 21. Jahrhundert" verabschiedet, doch muss diese Strategie in den Mitgliedstaaten nun mit Leben erfüllt werden.

Show publication details
Silva, Nelson; Schreck, Tobias; Veas, Eduardo; Sabon, Vedran; Eggeling, Eva; Fellner, Dieter W.

Leveraging Eye-gaze and Time-series Features to Predict User Interests and Build a Recommendation Model for Visual Analysis

2018

ETRA '18

ACM Symposium on Eye Tracking Research & Applications (ETRA) <10, 2018, Warsaw, Poland>

We developed a new concept to improve the efficiency of visual analysis through visual recommendations. It uses a novel eye-gaze based recommendation model that aids users in identifying interesting time-series patterns. Our model combines time-series features and eye-gaze interests, captured via an eye-tracker. Mouse selections are also considered. The system provides an overlay visualization with recommended patterns, and an eye-history graph, that supports the users in the data exploration process. We conducted an experiment with 5 tasks where 30 participants explored sensor data of a wind turbine. This work presents results on pre-attentive features, and discusses the precision/recall of our model in comparison to final selections made by users. Our model helps users to efficiently identify interesting time-series patterns.

  • 978-1-4503-5706-7
Show publication details

Lowering the Barrier for Successful Replication and Evaluation

2018

BELIV 2018: Evaluation and Beyond - Methodological Approaches for Visualization

Workshop on evaluation and BEyond - methodoLogIcal approaches for Visualization (BELIV) <7, 2018, Berlin, Germany>

Evaluation of a visualization technique is complex and time-consuming. We present a system that aims at easing design, creation and execution of controlled experiments for visualizations in the web. We include of parameterizable visualization generation services, thus separating the visualization implementation from study design and execution. This enables experimenters to design and run multiple experiments on the same visualization service in parallel, replicate experiments, and compare different visualization services quickly. The system supports the range from simple questionnaires to visualization-specific interaction techniques as well as automated task generation based on dynamic sampling of parameter spaces. We feature two examples to demonstrate our service-based approach. One example demonstrates how a suite of successive experiments can be conducted, while the other example includes an extended replication study.

  • 978-1-5386-6884-9
Show publication details
Riexinger, Günther; Kluth, Andreas; Olbrich, Manuel; Braun, Jan-Derrik; Bauernhansl, Thomas

Mixed Reality for On-Site Self-Instruction and Self-Inspection with Building Information Models

2018

Procedia CIRP

CIRP Conference on Manufacturing Systems (CIRP CMS) <51, 2018, Stockholm, Sweden>

Mixed Reality (MR) solutions for the manufacturing domain aim to support the construction, production and maintenance processes of factories, its equipment and products. Within the European project “INSITER”, Fraunhofer developed visualization solutions and MR prototypes aiming to provide relevant data for different stakeholders using Building Information Modeling (BIM)-based information. Planning data of production environments or construction sites is made available on site via MR. The main objective of the work presented in this paper, is to support planning processes as well as production and construction workflows with in-situ visualization of digital planning or process data in MR.

Show publication details

Mobile Assisted Living: Smartwatch-based Fall Risk Assessment for Elderly People

2018

iWOAR 2018

International Workshop on Sensor-based Activity Recognition (iWOAR) <5, 2018, Rostock, Germany>

ACM International Conference Proceedings Series

We present a novel Smartwatch-based approach, to enable Mobile Assisted Living (MAL) for users with special needs. A major focus group for this approach are elderly people. We developed a tool for caregivers applicable in home environments, nursing care, and hospitals, to assess the vitality of their patients. Hereby, we particularly focus on the prediction of falls, because falls are a major reason for serious injuries and premature death among elderly. Therefore, we propose a multi parametric score based on standardized fall risk assessment tests, as well as on sleep quality, medication, patient history, motor skills, and environmental factors. The resulting total fall risk score reflects individual changes in behavior and vitality, which consequently enables for fall preventing interventions. Our system has been deployed and evaluated in a pilot study among 30 elderly patients over a period of four weeks.

  • 978-1-4503-6487-4
Show publication details
Mirashi, Sudesh Ganapati; Glesner, Manfred [1. Gutachten]; Wirtz, Andreas [Betreuer]

Model-based Segmentation of the Teeth in Panoramic Radiograph Images

2018

Darmstadt, TU, Master Thesis, 2018

In this thesis, a fully automatic approach for teeth segmentation in dental panoramic radiograhpic images is presented. The approach uses an exsiting coupled shape model framework, which was developed at Fraunhofer IGD, in conjunction with a convolutional neural network (CNN). The CNN provides a preliminary segmentation of the teeth region which is used to initialize the coupled shape model in terms of position and scale. Then, the 28 individual teeth (excluding wisdom teeth) are segmented and labeled using gradient image features in combination with the statistical knowledge about their shape variation and spatial relation. A combination of separate adaptation steps is used to ensure a robust segmentation. The segmentation quality of the approach is assessed by comparing the generated results to manually created gold-standard segmentations of the individual teeth.

Show publication details
Damer, Naser; Moseguí Saladié, Alexandra; Braun, Andreas; Kuijper, Arjan

MorGAN: Recognition Vulnerability and Attack Detectability of Face Morphing Attacks Created by Generative Adversarial Network

2018

IEEE 9th International Conference on Biometrics: Theory, Applications and Systems

IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS) <9, 2018, Redondo Beach, CA, USA>

Face morphing attacks aim at creating face images that are verifiable to be the face of multiple identities, which can lead to building faulty identity links in operations like border crossing. Research has been focused on creating more accurate attack detection approaches by considering different image properties. However, all the attacks considered so far are based on manipulating facial landmarks localized in the morphed face images. In contrast, this work presents novel face morphing attacks based on image generated by generative adversarial networks. We present the MorGAN structure that considers the representation loss to successfully create realistic morphing attacks. Based on that, we present a novel face morphing attacks database (MorGAN database) that contains 1000 morph images for both, the proposed MorGAN and landmark-based attacks. We present vulnerability analysis of two face recognition approaches facing the proposed attacks. Moreover, the detectability of the proposed MorGAN attacks is studied, in the scenarios where this type of attacks is know and unknown. We concluded with pointing out the challenge of detecting such unknown novel attacks and an analysis of detection performances of different features in detecting such attacks.

  • 978-1-5386-7180-1
Show publication details

MoST - A 3DWeb Architectural Style for Hybrid Model Data

2018

Proceedings Web3D 2018

International Conference on 3D Web Technology (WEB3D) <23, 2018, Poznań, Poland>

Within this paper, we present a novel 3D web architecture style which allows to build format-agnostic 3D model graphs on the basis of ReSTful principles. We generalize the abstract definitions of RFC 2077 and allow to compose models and model fractions while transferring the "Media Selection URI" to a new domain. We present a best practice subset of HTTP/HTTPS and ReST to model authorization, data change, and content format negation within a single efficient request. This allows implementations to handle massive graphs with hybrid format configurations on the very efficient HTTP transport layer, without further application intervention. The system should be attractive to platform and service providers aiming to increase their ability to build 3D data application mashups with a much higher level of interoperability. We also hope to inspire standardization organizations to link generic "model/*" formats to RFC 2077 defined semantics like "compose".

  • 978-1-4503-5800-2
Show publication details
Behr, Johannes; Stein, Christian; Sousa, Miguel de

Non-invasive 3D Data Access for PLM Work Flows

2018

Proceedings Web3D 2018

International Conference on 3D Web Technology (WEB3D) <23, 2018, Poznań, Poland>

We present the instant3Dhub solution as a showcase on the use of today’s web technologies for connecting business data and 3D data in the context of PLM systems and modern service platforms or cloud environments. This solution allows companies to easily enrich their applications and workflows with 3D visualization in a non-invasive way. Concepts and benefits are illustrated by examples of successfully integrated solutions.

  • 978-1-4503-5800-2
Show publication details
Wulf, Conrad; Urban, Bodo [Betreuer]; Bieber, Gerald [Betreuer]

Optische Erfassung und Analyse von Pulswellen

2018

Rostock, Univ., Bachelor Thesis, 2018

Die vorliegende Bachelorarbeit untersucht den Ursprung periodischer Schwankungen im Remote-Photoplethysmographie-Signal. Dazu wurden Videoaufnahmen einer Person in zwei verschiedenen Haltungen angefertigt. Parallel dazu wurde ein photoplethysmographisches Signal an der Stirn und am Finger gemessen. Aus den Videoaufnahmen wurde ein Remote- Photoplethysmographie-Signal der Stirn, Bewegungen der Stirn und Bewegungen des Brustkorbs extrahiert. Die Signale wurden einer Fourier-Analyse unterzogen und verglichen. Die Ergebnisse legen nahe, dass atemsynchrone Schwankungen des Remote-Photoplethysmographie- Signals nicht durch Bewegungen, sondern durch Farbänderungen der Haut entstehen. Gleiches gilt für eine weitere beobachtete Schwankung, deren Ursprung in der Mayerwelle vermutet wird. Diese war nur an der Stirn messbar und hatte abhängig von der Haltung der untersuchten Person verschieden große Amplituden.

Show publication details
Stoyanov, Danail [Ed.]; Taylor, Zaike [Ed.]; Sarikaya, Duygu [Ed.]; McLeod, Jonathan [Ed.]; González Ballester, Miguel Angel [Ed.]; Codella, Noel C.F. [Ed.]; Martel, Anne [Ed.]; Maier-Hein, Lena [Ed.]; Malpani, Anand [Ed.]; Zenati, Marco A. [Ed.]; De Ribaupierre, Sandrine [Ed.]; Xiongbiao, Luo [Ed.]; Collins, Toby [Ed.]; Reichl, Tobias [Ed.]; Drechsler, Klaus [Ed.]; Erdt, Marius [Ed.]; Linguraru, Marius George [Ed.]; Oyarzun Laura, Cristina [Ed.]; Shekhar, Raj [Ed.]; Wesarg, Stefan [Ed.]; Celebi, M. Emre [Ed.]; Dana, Kristin [Ed.]; Halpern, Allan [Ed.]

OR 2.0 Context-Aware Operating Theaters, Computer Assisted Robotic Endoscopy, Clinical Image-Based Procedures, and Skin Image Analysis

2018

International Workshop on OR 2.0 Context-Aware Operating Theaters (OR 2.0) <1, 2018, Granada, Spain>

Lecture Notes in Computer Science (LNCS), Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP)
11041, 11041
  • 978-3-030-01200-7
Show publication details

P-score: Performance Aligned Normalization and an Evaluation in Score-level Multi-biometric Fusion

2018

2018 Proceedings of the 26th European Signal Processing Conference (EUSIPCO)

European Signal Processing Conference (EUSIPCO) <26, 2018, Rome, Italy>

Normalization is an important step for different fusion, classification, and decision making applications. Previous normalization approaches considered bringing values from different sources into a common range or distribution characteristics. In this work we propose a new normalization approach that transfers values into a normalized space where their relative performance in binary decision making is aligned across their whole range. Multi-biometric verification is a typical problem where information from different sources are normalized and fused to make a binary decision and therefore a good platform to evaluate the proposed normalization.We conducted an evaluation on two publicly available databases and showed that the normalization solution we are proposing consistently outperformed state-of-the-art and best practice approaches, e.g. by reducing the false rejection rate at 0.01% false acceptance rate by 60- 75% compared to the widely used z-score normalization under the sum-rule fusion.

  • 978-90-827970-1-5
Show publication details
Ritter, Christian; Altenhofen, Christian; Zeppelzauer, Matthias; Kuijper, Arjan; Schreck, Tobias; Bernard, Jürgen

Personalized Visual-Interactive Music Classification

2018

EuroVA 2018

International EuroVis Workshop on Visual Analytics (EuroVA) <9, 2018, Brno, Czech Republic>

We present an interactive visual music classification tool that will allow users to automatically structure music collections in a personalized way. With our approach, users play an active role in an iterative process of building classification models, using different interactive interfaces for labeling songs. The interactive tool conflates interfaces for the detailed analysis at different granularities, i.e., audio features, music songs, as well as classification results at a glance. Interactive labeling is provided with three complementary interfaces, combining model-centered and human-centered labeling-support principles. A clean visual design of the individual interfaces depicts complex model characteristics for experts, and indicates our work-in-progress towards the abilities of non-experts. The result of a preliminary usage scenario shows that, with our system, hardly any knowledge about machine learning is needed to create classification models of high accuracy with less than 50 labels.

  • 978-3-03868-064-2
Show publication details
Korn, Oliver; Bieber, Gerald; Fron, Christian

Perspectives on Social Robots. From the Historic Background to an Experts’ View on Future Developments

2018

Proceedings of the 11th International Conference on PErvasive Technologies Related to Assistive Environments

ACM International Conference on PErvasive Technologies Related to Assistive Environments (PETRA) <11, 2018, Corfu, Greece>

ACM International Conference Proceedings Series (ICPS)
01608

Social robots are robots interacting with humans not only in collaborative settings, but also in personal settings like domestic services and healthcare. Some social robots simulate feelings (companions) while others just help lifting (assistants). However, they often incite both fascination and fear: what abilities should social robots have and what should remain exclusive to humans? We provide a historical background on the development of robots and related machines (1), discuss examples of social robots (2) and present an expert study on their desired future abilities and applications (3) conducted within the Forum of the European Active and Assisted Living Programme (AAL). The findings indicate that most technologies required for the social robots’ emotion sensing are considered ready. For care robots, the experts approve health-related tasks like drawing blood while they prefer humans to do nursing tasks like washing. On a larger societal scale, the acceptance of social robots increases highly significantly with familiarity, making health robots and even military drones more acceptable than sex robots or child companion robots for childless couples. Accordingly, the acceptance of social robots seems to decrease with the level of face-to-face emotions involved.

  • 978-1-4503-6390-7
Show publication details
Riffnaller-Schiefer, Andreas; Augsdörfer, Ursula H.; Fellner, Dieter W.

Physics-based Deformation of Subdivision Surfaces for Shared Virtual Worlds

2018

Computers & Graphics

International Conference on Cyberworlds (CW) <2017, Chester, UK>

Creating immersive interactive virtual worlds not only require plausible visuals, but it is also important to allow the user to interact with the virtual scene in a natural way. While rigid-body physics simulations are widely used to provide basic interaction, realistic soft-body deformations of virtual objects are challenging and therefore typically not offered in multi user environments. We present a web service for interactive deformation which can accurately replicate real world material behavior. Its architecture is highly flexible, can be used from any web enabled client, and facilitates synchronization of computed deformations across multiple users and devices at different levels of detail.

Show publication details
Fauser, Johannes; Sakas, Georgios; Mukhopadhyay, Anirban

Planning Nonlinear Access Paths for Temporal Bone Surgery

2018

International Journal of Computer Assisted Radiology and Surgery

Purpose: Interventions at the otobasis operate in the narrow region of the temporal bone where several highly sensitive organs define obstacles with minimal clearance for surgical instruments. Nonlinear trajectories for potentialminimally invasive interventions can provide larger distances to risk structures and optimized orientations of surgical instruments, thus improving clinical outcomes when compared to existing linear approaches. In this paper, we present fast and accurate planning methods for such nonlinear access paths. Methods: We define a specific motion planning problem in SE(3) = R3 × SO(3) with notable constraints in computation time and goal pose that reflect the requirements of temporal bone surgery. We then present k-RRT-Connect: two suitable motion planners based on bidirectional Rapidly exploring Random Tree (RRT) to solve this problem efficiently. Results: The benefits of k-RRT-Connect are demonstrated on real CT data of patients. Their general performance is shown on a large set of realistic synthetic anatomies. We also show that these new algorithms outperform state-of-the-art methods based on circular arcs or Bézier-Splines when applied to this specific problem. Conclusion: With this work, we demonstrate that preoperative and intra-operative planning of nonlinear access paths is possible for minimally invasive surgeries at the otobasis.

Show publication details
Sacco, Luís Rüger; Kuijper, Arjan [1. Prüfer]; Siegmund, Dirk [2. Prüfer]

Predicting OCR Errors in Natural Scene Images

2018

Darmstadt, TU, Bachelor Thesis, 2018

This thesis explores the usage of Image Quality Assessment (IQA) Systems in order to increase the reliability of OCR systems in the natural scene. It proposes to increase the reliability of OCR in natural scene, based on the principle that OCR accuracy is a function of the quality of the input image. This work focus on assessing image quality from video frames in real time in order to pick high-quality images for the OCR process. The IQA system predicts OCR error chances and outputs the image quality problems, such as blur and light effects. The key technology developed for this work is an efficient IQA System built using the MobileNet V1 Convolutional Neural Network (CNN) Architecture. The approach behind the system builds upon past research on CNN based IQA and mainly on transfer learning research in the field done by Bianco et al. [1], which states that CNNs trained for general object recognition already learn important features for IQA tasks. The final system was pre-trained for object recognition using the ImageNet dataset and had the last fully connected layer retrained for the new IQA classification task using a database with 180k images created in this work. The database is divided into three classes with 60k images each. The "Good" class contains natural scene text boxes that can be read by the Tesseract engine with no errors. The "Light" class contain natural scene text boxes where light effects caused an OCR error and the "Blur" class contains natural scene text boxes where blur caused an OCR error. The CNN achieved a classification accuracy of 99.35% on the validation set containing 17k text boxes images and performs as expected in real natural scene text images.

Show publication details
Pammer-Schindler, Viktoria; Aehnelt, Mario; Klamma, Ralf

Preface: Human Computer Interaction Perspectives on Industry 4.0

2018

Interaction Design and Architecture(s) Journal - IxD&A
Show publication details
Rus, Silvia; Hammacher, Felix; Wilmsdorff, Julian von; Braun, Andreas; Große-Puppendahl, Tobias; Kirchbuchner, Florian; Kuijper, Arjan

Prototyping Shape-Sensing Fabrics Through Physical Simulation

2018

Ambient Intelligence

European Conference on Ambient Intelligence (AmI) <14, 2018, Larnaca, Cyprus>

Lecture Notes in Computer Science (LNCS)
11249

Embedding sensors into fabrics can leverage substantial improvements in application areas like working safety, 3D modeling or health-care, for example to recognize the risk of developing skin ulcers. Finding a suitable setup and sensor combination for a shape-sensing fabric currently relies on the intuition of an application engineer. We introduce a novel approach: Simulating the shape-sensing fabric first and optimize the design to achieve better real-world implementations. In order to enable developers to easily prototype their shape-sensing scenario, we have implemented a framework that enables soft body simulation and virtual prototyping. To evaluate our approach, we investigate the design of a system detecting sleeping postures. We simulate potential designs first, and implement a bed cover consisting of 40 distributed acceleration sensors. The validity of our framework is confirmed by comparing the simulated and real evaluation results. We show that both approaches achieve similar performances, with an F-measure of 85% for the virtual prototype and 89% for the real-world implementation.

  • 978-3-030-03061-2
Show publication details
Aehnelt, Mario; Bader, Sebastian

Providing and Adapting Information Assistance for Smart Assembly Stations

2018

Proceedings of SAI Intelligent Systems Conference (IntelliSys) 2016

SAI Intelligent Systems Conference (IntelliSys) <20, 2016, London, UK>

Information assistance helps in many application domains to structure, guide and control human work processes. However, it lacks a formalisation and automated processing of background knowledge which vice versa is required to provide ad-hoc assistance. In this paper, we describe our conceptual and technical work to provide information assistance for smart assembly stations. Our contribution comprises a conceptual architecture which was implemented in an industrial prototype, the Plant@Hand smart assembly trolley. We describe three major aspects of our approach, the recognition of assembly tasks based on probabilistic models, decision making using a cognitive architecture, and strategies for dealing with sensor-based and knowledge-based errors. Finally, we present the industrial prototype setup.

Show publication details
Behrisch, Michael; Blumenschein, M.; Kim, N. W.; Shao, Lin; El-Assady, M.; Fuchs, Johannes; Seebacher, Daniel; Diehl, Alexandra; Brandes, U.; Pfister, Hanspeter; Schreck, Tobias; Weiskopf, Daniel; Keim, Daniel A.

Quality Metrics for Information Visualization

2018

Computer Graphics Forum

Eurographics / IEEE VGTC Conference on Visualization (EuroVis) <20, 2018, Brno, Czech Republic>

The visualization community has developed to date many intuitions and understandings of how to judge the quality of views in visualizing data. The computation of a visualization’s quality and usefulness ranges from measuring clutter and overlap, up to the existence and perception of specific (visual) patterns. This survey attempts to report, categorize and unify the diverse understandings and aims to establish a common vocabulary that will enable a wide audience to understand their differences and subtleties. For this purpose, we present a commonly applicable quality metric formalization that should detail and relate all constituting parts of a quality metric. We organize our corpus of reviewed research papers along the data types established in the information visualization community: multi- and high-dimensional, relational, sequential, geospatial and text data. For each data type, we select the visualization subdomains in which quality metrics are an active research field and report their findings, reason on the underlying concepts, describe goals and outline the constraints and requirements. One central goal of this survey is to provide guidance on future research opportunities for the field and outline how different visualization communities could benefit from each other by applying or transferring knowledge to their respective subdomain. Additionally, we aim to motivate the visualization community to compare computed measures to the perception of humans.

Show publication details
Matthies, Denys J.C.; Urban, Bodo [Thesis Reviewer]; Schmidt, Albrecht [Thesis Reviewer]; Wolf, Katrin [Thesis Reviewer]

Reflexive Interaction - Extending Peripheral Interaction by Augmenting Humans

2018

Rostock, Univ., Diss., 2018

Technology is closer to the human than ever, it exists in various shapes and forms, is omnipresent, while continuously competing for the user's attention. With new opportunities constantly arising, such as mobile computing, we also face challenges, particularly when the user is on the go. Because of mobile devices often demand the user's full attention, control in mobile scenarios can be complicated, inadequate, awkward, risky, or not feasible at all. To overcome these problems, the concept of a Reflexive Interaction is presented, which can be seen as a specific manifestation of Peripheral Interaction. In contrast, a Reflexive Interaction is envisioned to be executed at a secondary task without involving substantial cognitive effort, while enabling the user tiny interactions, shorter than Microinteractions, without straining the user's main interaction channels occupied with the primary task. To underline the proposed concept, a series of research studies has been conducted that exploit the unique sensing and motor capabilities of the human body. For this, three body regions (head, body, and foot) have been selected, which all yield specific characteristics. For instance, the region of the head enables facial gesture control, while visual information is perceivable within our peripheral vision. On our body, quick tapping and hovering can be performed, while haptic, thermal, or electrical feedback can be applied on our skin in order to perceive different scales of notifications. The foot enables quick foot tapping gestures as well as the possibility to perceive vibrotactile feedback under the foot's sole. Moreover, in particular the foot, but also the face, generates unique information, which can be utilized to infer on the user's context, such as physical activity or emotional state. The consideration of context information is important in order to determine whether and how a Reflexive Interaction can be implemented.

Show publication details
Albadawi, Mohamad; Lukas, Uwe von [Supervising professor]; Krause, Tom [Tutor]

Resolving Classification Ambiguities in Convolutional Neural Networks Using Hierarchical Structures

2018

Rostock, Univ., Master Thesis, 2018

We have recently witnessed the revolution of deep learning and convolutional neural networks enabled by the powerful machines available today. Convolutional neural networks have demonstrated excellent performance on various vision tasks, most importantly classification and detection. Nevertheless, there are some difficulties in the way of perfect performance. One problem is discriminating among objects that look extremely similar visually but semantically they are different. Another problem is the high cost of training large detection models. The same cost applies when the model is required to detect a new type of objects. In this work those problems are handled by introducing visual concepts and the use of hierarchical structures. We will see how the accuracy of classifying similar objects can be highly improved and how the time of accommodating for new objects in a detection model can be reduced from days to hours.

Show publication details
Tang, Chong; Wang, Rui; Wang, Yu; Wang, Shuo; Lukas, Uwe von; Tan, Min

RobCutt: A Framework of Underwater Biomimetic Vehicle-Manipulator System for Autonomous Interventions

2018

2018 14th IEEE International Conference on Automation Science and Engineering (CASE)

IEEE International Conference on Automation Science and Engineering (CASE) <14, 2018, Munich, Germany>

This paper presents a general concept framework of the underwater biomimetic vehicle-manipulator system (UBVMS) for autonomous interventions in terms of objectives, as well as technologies and methodologies. With full consideration of the autonomous cruise and intervention, the RobCutt system’s configuration and methodology are designed to promote the levels of autonomy of the autonomous underwater vehiclemanipulator system (UVMS). The second generation UBVMS (RobCutt II) is introduced, including the design and principle of the biomimetic propulsor inspired by the cuttlefish and lightweight manipulator, and the advantages are concluded. Moreover, technologies and methodologies of underwater localization, object detection and coordination control are designed and accomplished respectively. Finally, pool tests have been carried out to verify the feasibility and effectiveness of the developed framework and methodology.

  • 978-1-5386-2514-9
Show publication details
Matthies, Denys J.C.; Daza Parra, Laura Milena; Urban, Bodo

Scaling Notifications Beyond Alerts: From Subtly Drawing Attention up to Forcing the User to Take Action

2018

UIST 2018 Adjunct

Research has been done in sophisticated notifications, still, devices today mainly stick to a binary level of information, while they are either attention drawing or silent. We propose scalable notifications, which adjust the intensity level reaching from subtle to obtrusive and even going beyond that level while forcing the user to take action. To illustrate the technical feasibility and validity of this concept, we developed three prototypes. The prototypes provided mechano-pressure, thermal, and electrical feedback, which were evaluated in different lab studies. Our first prototype provides subtle poking through to high and frequent pressure on the user’s spine, which significantly improves back posture. In a second scenario, the user is able to perceive the overuse of a drill by an increased temperature on the palm of a hand until the heat is intolerable, forcing the user to eventually put down the tool. The last application comprises of a speed control in a driving simulation, while electric muscle stimulation on the users’ legs, conveys information on changing the car’s speed by a perceived tingling until the system forces the foot to move involuntarily. In conclusion, all studies’ findings support the feasibility of our concept of a scalable notification system, including the system forcing an intervention.

  • 978-1-4503-5949-8
Show publication details
Schinko, Christoph; Fellner, Dieter W. [Supervisor]; Klein, Reinhard [Supervisor]

Shape Processing for Content Generation

2018

Graz, TU, Diss., 2018

In seiner Dissertation „Shape Processing for Content Generation“ behandelt Christoph Schinko generative Modellierung, neuartige Anwendungen der inversen generativen Modellierung sowie Visualisierungssysteme. Diese Bereiche werden als Bestandteile der Formverarbeitung betrachtet, daher richtet sich der Aufbau der Dissertation danach. Nach der Definition des Begriffs " Form\\ befasst sich der erste Teil der Dissertation mit verschiedenen Möglichkeiten der Formbeschreibung. Während einige Formbeschreibungen von abstrakter Natur sind, können andere direkt verwendet werden - beispielsweise auf dem Gebiet der computergestützten geometrischen Gestaltung. Das Thema "Formmodellierung“ (kurz: Modellierung) ist breit gefächert und umfasst die Modellierung mit Primitiven unter Zuhilfenahme von 3D-Modellierungssoftware oder Szenenbeschreibungssprachen, semantische Modellierung mit Metadaten sowie generative Modellierung mit domänenspezifischen Informationen. Am Beispiel von Trauringen wird mit Hilfe der Generative Modeling Language (GML), einer domänenspezifischen Sprache für die generative Modellierung, ein Design einer ganzen Produktfamilie erstellt. Bei der Auslieferung des Designs über das Web sind unterschiedlichste Plattformen beteiligt. Dieser Umstand lieferte die Idee zu einem innovativen Metamodellier-Ansatz namens „Euclides“. Das innovative Konzept kombiniert die Unterstützung verschiedener Zielplattformen mit einer anfängerfreundlichen Syntax. Damit wird die Grundlage für die plattformunabhängige Generierung von generativen Bausteinen geschaffen. Dieser Ansatz reduziert den Aufwand für die Implementierung und Pflege generativer Beschreibungen für verschiedene Plattformen erheblich. Aufbauend auf Arbeiten der inversen generativen Modellierung wird die Analyse von digitalisierten Objekten hinsichtlich Veränderungen und Abnutzung möglich. Das vorgestellte System kombiniert generative Beschreibungen mit rekonstruierten Objekten und führt einen Soll-Ist-Wert-Vergleich durch. Durch Anwendung auf einen anderen Parametersatz der generativen Beschreibung können somit neue Formen erzeugt werden. Mit diesem neuartigen Ansatz ist die Gestaltung von Formen unter gleichzeitiger Verwendung hochfrequenter Details sowie High-Level-Form-Parametern möglich. Der letzte Schritt im Rahmen der Formverarbeitung befasst sich mit Visualisierungssystemen zur Wahrnehmung und Interaktion mit Formen. In diesem Zusammenhang wird eine neuartige Methode zur Projektion eines kohärenten, nahtlosen und perspektivisch korrigierten Bildes von einem bestimmten Gesichtspunkt aus vorgestellt. Der Ansatz zeichnet sich vor allem durch seine Effizienz aus. Der letzte Beitrag zu diesem Thema beschreibt eine optimierte, autostereoskopische Visualisierung auf Basis von Parallaxbarrieren.

Show publication details
Metko, Frederik; Weber, Hans-Peter [Referent]; Hergenröther, Elke [Korreferentin]; Tausch, Reimar [Betreuer]

Silhouetten-Schnittverfahren unterstützte Ansichtenplanung zur automatisierten 3D-Digitalisierung

2018

Darmstadt, Hochschule, Bachelor Thesis, 2018

Die Digitalisierung von Objekten ist zeitaufwendig und mit sehr viel manueller Arbeit verbunden. Soll dieser Vorgang automatisiert werden, wird ein virtuelles 3D Modell des zu scannenden Objektes benötigt um den Scanvorgang zu planen. Wie soll allerdings vorgegangen werden, wenn ein Objekt noch nie digital erfasst wurde und aus diesem Grund ein solches Modell nicht zur Verfügung steht? In dieser Arbeit wird ein Verfahren vorgestellt, das es ermöglicht auch unbekannte Objekte mittels photogrammetrischen Verfahrens automatisch zu digitalisieren. Dazu wird mit Hilfe des Silhouetten-Schnittverfahrens das benötigte virtuelle Modell erzeugt und darauf basierend die Erfassung des zu scannenden Objektes mit einer Ansichtenplanung geplant. Aus diesem Verfahren wurde ein System entwickelt, mit dem es möglich ist, beliebige Objekte mit minimaler Benutzerinteraktion automatisch zu digitalisieren.

Show publication details
Merz, Johannes; Getto, Roman; Kuijper, Arjan; Fellner, Dieter W.

Simplified Definition of Parameter Spaces of a Procedural Model using Sketch-based Interaction

2018

GRAPP 2018. Proceedings

International Conference on Computer Graphics Theory and Applications (GRAPP) <13, 2018, Funchal, Madeira, Portugall>

This paper presents a novel technique to intuitively insert meta-parameters into a procedural model with the help of sketch-based interaction. The procedural model is represented in a GML (Generative Modeling Language) representation, which is a script language that focuses on the description of three-dimensional models. A GML model consists of a sequence of procedural modeling commands, for example extrusions. These are called with a set of local offset positions, which can be converted to global space and anchored in the surface mesh by finding reference vertices on the mesh. The system uses a deformation technique to deform the surface of the model. During the deformation, the reference vertices provide the global offset positions, whose path can be approximated by a B-spline. Exchanging the initial values of the commands by this B-spline, defines a continuous parameter space of the meta-parameter. The deformation process is supported by a mesh segmentation to create pre-defined deformation targets. Using sketch-based methods, these can be adapted to the user's needs. The results show that the system closely imitates the deformation with the help of the modeling commands. Furthermore, the system was evaluated to be intuitive and easy to use.

Show publication details
Seidel, Viviane; Stork, André [1. Supervisor]; Weber, Daniel [2. Supervisor]

Simulation of 4D Programmable Textiles created by 3D Print on Prestressed Fabric

2018

Darmstadt, TU, Bachelor Thesis, 2018

The objective of this thesis is to develop a simulation with the finite element method for three-dimensional imprinted, prestretched textile. For this purpose, methods necessary to model the intended behavior are conceived and tested. More specifically, methods are developed that should move the imprinted elements towards the stretched state, in which they have been when they got imprinted. A simple adjustment of the rest positions of the vertices as well as a modification of the derivative of the basis functions were found out not to be expedient. The application of an initial strain was discovered to be a functional strategy to maintain the stretched state. Adding a second layer of finite elements to the imprinted sections turned out to be a viable option to integrate the two different material characteristics of the components. However, this approach does not keep up the stretch of the imprinted areas, since the elements are still only connecting the same old vertices with the old initial rest states. Finally two options for the required perturbation of the simulation are introduced and their impact is analyzed. The combination of the respectively selected approaches results in a realistic simulation generating a visually convincing reproduction of the reality of 4D programmable textiles. Moreover, the linear model allows for an interactive modification of the resulting shape by modifying the scenario and adjusting material parameters.

Show publication details
Hüther, Janina; Ben Hmida, Helmi [1. Gutachten]; Kuijper, Arjan [2. Gutachten]

Smart Recommendation Systems for IoT

2018

Darmstadt, TU, Master Thesis, 2018

Nowadays Internet of Things is a important research field in Computer Science. The personalization of IoT services is important for the usability and user experience, but the rapidly growing number of IoT services and therefore the increasing number of possibilities, makes it hard for the user to configure an IoT system. Recommender systems are there to automate the decision process. But a lot of different recommender algorithms exist and it is hard to decide for a certain method regarding a specific IoT Use Case. In this master thesis recommender system methods are investigated in regard to their characteristics as well as IoT properties and scenarios. The goal of this research is to suggest a generic model for selecting and executing recommender system methods depending on a given IoT scenario. All defined tasks were accomplished successfully. A complete concept was created and realized as an web interface application, which was evaluated with data of two IoT use cases.

Show publication details
Sacha, Dominik; Kraus, Matthias; Bernard, Jürgen; Behrisch, Michael; Schreck, Tobias; Asano, Yuki; Keim, Daniel A.

SOMFlow: Guided Exploratory Cluster Analysis with Self-Organizing Maps and Analytic Provenance

2018

IEEE Transactions on Visualization and Computer Graphics

Clustering is a core building block for data analysis, aiming to extract otherwise hidden structures and relations from raw datasets, such as particular groups that can be effectively related, compared, and interpreted. A plethora of visual-interactive cluster analysis techniques has been proposed to date, however, arriving at useful clusterings often requires several rounds of user interactions to fine-tune the data preprocessing and algorithms. We present a multi-stage Visual Analytics (VA) approach for iterative cluster refinement together with an implementation (SOMFlow) that uses Self-Organizing Maps (SOM) to analyze time series data. It supports exploration by offering the analyst a visual platform to analyze intermediate results, adapt the underlying computations, iteratively partition the data, and to reflect previous analytical activities. The history of previous decisions is explicitly visualized within a flow graph, allowing to compare earlier cluster refinements and to explore relations. We further leverage quality and interestingness measures to guide the analyst in the discovery of useful patterns, relations, and data partitions. We conducted two pair analytics experiments together with a subject matter expert in speech intonation research to demonstrate that the approach is effective for interactive data analysis, supporting enhanced understanding of clustering results as well as the interactive process itself.

Show publication details
Singh, Shashank; Kuijper, Arjan [1. Gutachter]; Siegmund, Dirk [2. Gutachter]

Speech Emotion Recognition as a Wearable Device for Depressed People

2018

Darmstadt, TU, Master Thesis, Jahr

Recognition of Emotions from speech was first coined by the works of Daellert et.al in 1996[7], published as the first research paper regarding this topic. It was re-introduced with the concepts of Deep Learning around 10 year back and has been in evolution since the last 2 decades. Communicating with machines have become a hot topic for research and industrial purposes, already seen quite a lot of advancements such as the once found on our smartphones: Cortana, Siri, Google Now to the very sophisticated Artificially Intelligent "Sophia" [35] robot, with the heart of technology centered around predicting human emotions. This field of automatically recognizing human emotions and affective (mental) states also know commonly as Speech Emotion Recognition (SER) has seen an amalgamation of different technologies thereby blurring the line between SER and Artificial Intelligence. Predicting emotions in a human being, is a challenging task for machines, let alone even we as humans sometimes fail to predict the correct emotions in other person. The advancements in DL have led to many breakthroughs in this field and have come a long way from predicting basic emotions of happiness, anger, sad, fear as explained by Paul Ekman to predicting real time emotions. The real time emotions can be defined as the spontaneous ones which show variations with time as is the case with the real human emotions. A human being exhibits various affective states i.e. emotions in his/her speech, governed by the various external and internal factors. The idea behind this thesis is to predict such spontaneous emotions, in two emotional dimensions called the Arousal- which is said to be accessible to acoustic features such as the vibrations made by the vocal chords in a person. Valence-said to be accessible to the linguistic features such as those particular to a language or linguistic [40]. Also, these two dimensions are correlated and could be thought of as the coordinates of a particular emotion on a graph plot against Time, for example happiness can be translated to (positive valence, high (positive) arousal). Emotions are not constant and change with time, place and person, in other words, they have a contextual nature and therefore this thesis attempts to make this study and provide the results. Depression has been a major and very common mental disease in today’s world. Various factors such as loneliness, stress, family loss could be, among many others potential causes. To predict the mental health of a depressed person using such a SER system is quite a challenging task and this thesis attempts to predict spontaneous emotions by adapting a recurrent convolutional DL speech architecture, in the two emotional dimensions. The acoustic features are extracted using the CNN layers, and the temporal structure of the speech is modelled using the LSTM layers i.e. providing the "contextual information" from the audio signal. And evaluate those continuous quantities using the Regression metrics of DL based on their correlations. The outcome of this thesis is to check the values of the metrics based on the valence, arousal dimensions which are predicted by the network, for a continuous quantity of audio frames (chunks of 40ms) provided as training data. And the corresponding annotations (for each of 40ms audio frames) is also prepared manually for each of the audio files. Concordance Correlation Coefficient is used as the metric to evaluate on the factors such as the covariance, mean etc. indicating the correlations between the arousal, valence predictions made by the network.

Show publication details

Step by Step: Early Detection of Diseases Using an Intelligent Floor

2018

Ambient Intelligence

European Conference on Ambient Intelligence (AmI) <14, 2018, Larnaca, Cyprus>

Lecture Notes in Computer Science (LNCS)
11249

The development of sensor technologies in smart homes helps to increase user comfort or to create safety through the recognition of emergency situations. For example, lighting in the home can be controlled or an emergency call can be triggered if sensors hidden in the floor detect a fall of a person. It makes sense to also use these technologies regarding prevention and early detection of diseases. By detecting deviations and behavioral changes through long-term monitoring of daily life activities it is possible to identify physical or cognitive diseases. In this work, we first examine in detail the existing possibilities to recognize the activities of daily life and the capability of such a system to conclude from the given data on illnesses. Then we propose a model for the use of floor-based sensor technology to help diagnose diseases and behavioral changes by analyzing the time spent in bed as well as the walking speed of users. Finally, we show that the system can be used in a real environment.

  • 978-3-030-03061-2
Show publication details

Step Detection for Rollator Users with Smartwatches

2018

SUI '18 Proceedings of the Symposium on Spatial User Interaction

ACM Symposium on Spatial User Interaction (SUI) <6, 2018, Berlin, Germany>

Smartwatches enable spatial user input, namely for the continuous tracking of physical activity and relevant health parameters. Additionally, smartwatches are experiencing greater social acceptability, even among the elderly. While step counting is an essential parameter to calculate the user’s spatial activity, current detection algorithms are insufficient for calculating steps when using a rollator, which is a common walking aid for elderly people. Through a pilot study conducted with eight different wrist-worn smart devices, an overall recognition of ~10% was achieved. This is because characteristic motions utilized by step counting algorithms are poorly reflected at the user’s wrist when pushing a rollator. This issue is also present among other spatial activities such as pushing a pram, a bike, and a shopping cart. This paper thus introduces an improved step counting algorithm for wrist-worn accelerometers. This new algorithm was first evaluated through a controlled study and achieved promising results with an overall recognition of ~85%. As a follow-up, a preliminary field study with randomly selected elderly people who used rollators resulted in similar detection rates of ~83%. To conclude, this research will expectantly contribute to greater step counting precision in smart wearable technology.

  • 978-1-4503-5708-1
Show publication details

Surface Acoustic Arrays to Analyze Human Activities in Smart Environments

2018

Ambient Intelligence

European Conference on Ambient Intelligence (AmI) <14, 2018, Larnaca, Cyprus>

Smart Environments should be able to understand a user’s need without explicit interaction. In order to do that, one step is to build a system that is able to recognize and track some common activities of the user. This way, we can provide a system that provides various services for controlling installed appliances and offering help for every day activities. Applying these services in the users’ environment should make his life more comfortable, easier, and safer. In this paper, we will introduce an embedded sensor system using surface acoustic arrays to analyze human activities in a smart environment. We divided basic activity groups ranging from walking, cupboard closing to falling, including their extended sub-activity groups. We expanded walking into walking barefoot, with shoes and with high heels and further extended closing cupboard with three cupboards locating on different positions. We further investigated the usage of single pickup or a combination of 4 pickups with their effect on the recognition precision. We achieved an overall precision of 97.23% with 10-fold cross validation using support vector machine (SVM) for all sub-activity group combined. Even using one pickup only, we can achieve an overall precision of more than 93%, but we can further increase the precision by using a combination of pickups up to 97.23%.

  • 978-3-030-03061-2
Show publication details

SurfaceVox - Exploring Sound Control for Gesture-Tracking Interactive Surfaces

2018

14th International Conference on Signal-Image Technology & Internet-Based Systems

International Conference on Signal Image Technology & Internet-Based Systems (SITIS) <14, 2018, Las Palmas de Gran Canaria, Spain>

Almost 100 years ago, the thereminvox was the first electronic musical instrument that could be controlled without contact. With precise positioning of two hands, the player controls pitch and volume of a sine sound, by changing the distance from two antennas. We present SurfaceVox, which combines the technology behind the thereminvox with an additional acoustic sensor to create a musical instrument that combines mid-air and touch gesture control. We explore various scenarios of sound synthesis and combine the system with an augmented reality application. SurfaceVox has been evaluated in a study with thirteen users for input precision, perceived workload, as well as pragmatic and hedonistic qualities of the application.

  • 978-1-5386-9385-8
Show publication details

Synthesis and Rendering of Seamless and Non-Repetitive 4D Texture Variations for Measured Optical Material Properties

2018

SIGGRAPH Asia 2018 Technical Briefs

Conference and Exhibition on Computer Graphics and Interactive Techniques in Asia (SIGGRAPH ASIA) <11, 2018, Tokyo, Japan>

We have lifted the one weakness of an existing fully automatic acquisition system for spatially varying optical material behavior of real object surfaces. While its expression of spatially varying material behavior with spherical dependence on incoming light as 4D texture (ABTF material model) allows flexible mapping on arbitrary 3D geometries, photo-realistic rendering and interaction in real-time, this very method of texture-like representation exposed it to common problems of texturing, striking in two levels. First, non-seamless textures create visible border artifacts. Second, even a perfectly seamless texture causes repetition artifacts due to side-by-side distribution in large numbers over the 3D surface. We solved both problems through our novel texture synthesis that generates a set of seamless texture variations randomly distributed on the surface at shading time. When compared to regular 2D textures, the inter-dimensional coherence of the 4D ABTF material model poses entirely new challenges to texture synthesis, which includes maintaining the consistency of material behavior throughout the space spanned by the spatial image domain and the angular illumination hemisphere. In addition, we tackle the increased memory consumption caused by the numerous variations through a fitting scheme specifically designed to reconstruct the most prominent effects captured in the material model.

  • 978-1-4503-6062-3
Show publication details
Siegmund, Dirk; Wainakh, Aidmar; Ebert, Tina; Braun, Andreas; Kuijper, Arjan

Text Localization in Born-Digital Images of Advertisements

2018

Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications

Iberoamerican Conference on Pattern Recognition (CIARP) <22, 2017, Valparaíso, Chile>

Localizing text in images is an important step in a number of applications and fundamental for optical character recognition. While born-digital text localization might look similar to other complex tasks in this field, it has certain distinct characteristics. Our novel approach combines individual strengths of the commonly used methods: stroke width transform and extremal regions and combines them with a method based on edge-based morphologically growing. We present a parameterfree method with high flexibility to varying text sizes and colorful image elements. We evaluate our method on a novel image database of different retail prospects, containing textual product information. Our results show a higher f-score than competitive methods on that particular task.

Show publication details
Samartzidis, Timotheos; Siegmund, Dirk; Gödde, Michael; Damer, Naser; Braun, Andreas; Kuijper, Arjan

The Dark Side of the Face: Exploring the Ultraviolet Spectrum for Face Biometrics

2018

2018 International Conference on Biometrics (ICB)

IAPR International Conference on Biometrics (ICB) <11, 2018, Gold Coast, Australia>

Facial recognition in the visible spectrum is a widelyused application but it is also still a major field of research.In this paper we present melanin face pigmentation (MFP)as a new modality to be used to extend classical face biometrics. Melanin pigmentation are sun-damaged cells thatoccur as revealed and/or unrevealed pattern on human skin.Most MFP can be found in the faces of some people whenusing ultraviolet (UV) imaging. To proof the relevance ofthis feature for biometrics, we present a novel image datasetof 91 multiethnic subjects in both, the visible and the UVspectrum. We show a method to extract the MFP featuresfrom the UV images, using the well known SURF featuresand compare it with other techniques. In order to proof itsbenefits, we use weighted score-level fusion and evaluatethe performance in an one against all comparison. As a resultwe observed a significant amplification of performancewhere traditional face recognition in the visible spectrum isextended with MFP from UV images. We conclude with afuture perspective about the use of these features for futureresearch and discuss observed issues and limitations.

Show publication details
Rus, Silvia; Joshi, Dhanashree Jayant; Braun, Andreas; Kuijper, Arjan [Betreuer]

The Emotive Couch - Learning Emotions by Capacitively Sensed Movements

2018

Procedia Computer Science [online]

International Conference on Ambient Systems, Networks and Technologies (ANT) <9, 2018, Porto, Portugal>

Affective computing allows machines to simulate and detect emotional states. The most common method is the observation of the face by camera. However, in our increasingly observed society, more privacy-aware methods are worth exploring that do not require facial images, but instead look at other physiological indicators of emotion. In this work we present the Emotive Couch, a sensor-augmented piece of smart furniture that detects proximity and motion of the human body. We present the design rationale and use standard machine learning techniques to detect the three basic emotions Anxiety, Interest, and Relaxation. We evaluate the performance of our approach with 15 participants in a study that includes various affect elicitation methods, achieving an accuracy of 77.7 %.

Show publication details
Bieber, Gerald; Antony, Niklas; Haescher, Marian

Touchless Heart Rate Recognition by Robots to Support Natural Human-Robot Communication

2018

Proceedings of the 11th International Conference on PErvasive Technologies Related to Assistive Environments

ACM International Conference on PErvasive Technologies Related to Assistive Environments (PETRA) <11, 2018, Corfu, Greece>

With the proliferation of robotic assistants, such as robot vacuum cleaners, telepresence robots, or shopping assistance robots, human-robot interaction becomes increasingly more natural. The capabilities of robots are expanding, which leads to an increasing need for a natural human-robot communication and interaction. Therefore, the modalities of text- or speech-based communication have to be extended by body language and a direct feedback such as emotion or non-verbal communication. In this paper, we present a camera-based, non-body contact optical heart rate recognition method that can be used in robots in order to identify humans' reactions during a robot-human communication or interaction. For the purpose of heart rate and heart rate variability detection, we used standard cameras (webcams) that are located inside the robot's eye. Although camera-based vital sign identification has been discussed in previous research, we noticed that certain limitations with regard to real-world applications do still exist. We identified artificial light sources as one of the main influencing factors. Therefore, we propose strategies with the aim Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from of improving natural communication between social robots and humans.

Show publication details
Bernard, Jürgen; Zeppelzauer, Matthias; Lehmann, Markus; Müller, Martin; Sedlmair, Michael

Towards User-Centered Active Learning Algorithms

2018

Computer Graphics Forum

Eurographics / IEEE VGTC Conference on Visualization (EuroVis) <20, 2018, Brno, Czech Republic>

The labeling of data sets is a time-consuming task, which is, however, an important prerequisite for machine learning and visual analytics. Visual-interactive labeling (VIAL) provides users an active role in the process of labeling, with the goal to combine the potentials of humans and machines to make labeling more efficient. Recent experiments showed that users apply different strategies when selecting instances for labeling with visual-interactive interfaces. In this paper, we contribute a systematic quantitative analysis of such user strategies. We identify computational building blocks of user strategies, formalize them, and investigate their potentials for different machine learning tasks in systematic experiments. The core insights of our experiments are as follows. First, we identified that particular user strategies can be used to considerably mitigate the bootstrap (cold start) problem in early labeling phases. Second, we observed that they have the potential to outperform existing active learning strategies in later phases. Third, we analyzed the identified core building blocks, which can serve as the basis for novel selection strategies. Overall, we observed that data-based user strategies (clusters, dense areas) work considerably well in early phases, while model-based user strategies (e.g., class separation) perform better during later phases. The insights gained from this work can be applied to develop novel active learning approaches as well as to better guide users in visual interactive labeling.

Show publication details
Farhadifard, Fahimeh; Urban, Bodo [Examiner]; Lukas, Uwe von [Betreuer]; Koch, Reinhard [Betreuer]

Underwater Image Restoration: Super-resolution and Deblurring via Sparse Representation and Denoising by Means of Marine Snow Removal

2018

Rostock, Univ., Diss., 2018

Underwater imaging has been widely used as a tool in many fields, such as marine industry, deep-sea mining, aquaculture and water assessment. However, a major issue is the quality of the resulting images and videos. Due to the light’s interaction with water and its constituents, the acquired underwater images and videos often suffer from a significant amount of scatter (blur and haze) and noise. Furthermore, since data transmission from the equipment mounted under water to the station above water is still a challenge, usually a compressed and low-resolution version of the data is transferred. In the light of these issues, this thesis considers the problems of low-resolution, blurred and noisy underwater images and proposes several approaches to improve the quality of such images/video frames. This is undertaken through two main contributions. The first major contribution of this work is the super-resolution and deblurring of single underwater images. This is done by using a set of compact high and low-resolution cluster dictionaries where sparse representation is used as the regularizer. Since such an approach inevitably calls for a model selection criterion in both learning and reconstruction stages, a scaleinvariance model is proposed to properly establish the link between the low and high-resolution feature spaces. The subject of the second major contribution is image denoising. Besides additive noises such as sensor noise, the visibility in underwater images is reduced by the presence of suspended particles in water. This represents an unwanted signal, which is also disruptive for advanced computer vision tasks, such as segmentation. Since this phenomenon is a real signal and part of the scene, two-fold approaches consisting of first detection and then removal of such particles, are proposed. To avoid the uncertainty introduced by using local information for restoration, some global priors of the scene are learned, which are then used to estimate the parts of the scene that are covered by the particles. For this, a Gaussian-based background subtraction approach is proposed to obtain static features of the scene. These are used as training data for learning the priors. Quantitative and qualitative experiments conducted over real and simulated underwater images and video frames validate the success of the proposed approaches at improving the image resolution and deblurring image features significantly as well as detecting and removing marine particles, while the object edges are preserved.

Show publication details
Meister, Simon; Hur, Junhwa; Roth, Stefan

UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss

2018

32nd AAAI Conference on Artificial Intelligence

AAAI Conference on Artificial Intelligence <32, 2018, New Orleans, Louisiana, USA>

In the era of end-to-end deep learning, many advances in computer vision are driven by large amounts of labeled data. In the optical flow setting, however, obtaining dense per-pixel ground truth for real scenes is difficult and thus such data is rare. Therefore, recent end-to-end convolutional networks for optical flow rely on synthetic datasets for supervision, but the domain mismatch between training and test scenarios continues to be a challenge. Inspired by classical energy-based optical flow methods, we design an unsupervised loss based on occlusion-aware bidirectional flow estimation and the robust census transform to circumvent the need for ground truth flow. On the KITTI benchmarks, our unsupervised approach outperforms previous unsupervised deep networks by a large margin, and is even more accurate than similar supervised methods trained on synthetic datasets alone. By optionally fine-tuning on the KITTI training data, our method achieves competitive optical flow accuracy on the KITTI 2012 and 2015 benchmarks, thus in addition enabling generic pre-training of supervised networks for datasets with limited amounts of ground truth.

  • 978-1-57735-800-8
Show publication details
Beuth, Marcel; Kuijper, Arjan [1. Gutachten]; Lücke-Tieke, Hendrik [2. Gutachten]

Usable Active Learner

2018

Darmstadt, TU, Master Thesis, 2018

Im Bereich des Machine Learnings ist eine der schwierigsten Aufgaben, einen geeigneten Trainingsdatensatz zu erstellen. Eines der bekanntesten Verfahren, um den Aufwand zu minimieren, welcher bei der Erzeugung eines solchen Trainingssets aufzuwenden ist, sind die sogenannten Active Learner. Diese Verfahren erstellen das Trainingsset in Kooperation mit einem Orakel. Bis heute existieren eine Vielzahl von verschiedenen Active-Learning-Verfahren. Diese verwenden verschiedene Metriken um die Datenpunkte zu selektieren, welche von dem Orakel gelabelt werden sollen. Zumeist wird bei diesen aber nur eine einzige Metrik verwendet. Daher befasst sich diese Arbeit mit einen Active-Learning-Verfahren, welches eine Kostenfunktion aus einer Vielzahl von verschiedenen Metriken konstruiert. Des Weiteren wird eine Webanwendung vorgestellt, welche dem Nutzer ermöglicht Korpusse und Modelle anzulegen. Zudem beinhaltet diese Webanwendung verschiedene Arten von Ansichten, mit deren Hilfe dem Nutzer das Labeln der Datenpunkte erleichtert werden soll. In dieser Arbeit werden zudem verschiedene Experimente durchgeführt, welche zum einem die Robustheit des Active-Learners belegen und zum anderen eine möglichst allgemeingültige Konfiguration der Freiheitsgrade definieren sollen.

Show publication details
Kannanayikkal, Clindo Devassy; Ulmer, Alex [Advisor]; Kuijper, Arjan [Supervisor]

User-Centered Anomaly Detection in Network Data

2018

Darmstadt, TU, Master Thesis, 2018

Identifying anomalies in network traffic logs is a very challenging task for a network analyst. With the ever-increasing number of devices that can be connected to the network, the need for detecting anomalies is at the peak. Usual techniques for detecting such anomalies include visual analysis of network data or applying automated algorithms. Both techniques have major drawbacks. Visual analysis requires high expertise of the analyst, and automated detection algorithms produce high rates of false alarms. In this work, both techniques are combined to improve the detection and reduce the workload of the analyst. The visual interface gives the network administrator the power to edit the predictions made by the algorithms. The feedback from the network administrator are used by the algorithms to improve the performance of the detector and to reduce the false alarms. The system is tested and evaluated on a publicly available dataset which shows that the system achieves competitive performance.

Show publication details
Dieter, Michelle; Kohlhammer, Jörn [1. Gutachten]; Gonzalez Parra, Garoe [2. Gutachten]

User-Centered Design of a Facilitator Tool for Pre-Processing and Visualizing Large Amounts of NOTAMs

2018

Darmstadt, TU, Master Thesis, 2018

NOTAMs (Notices to Airmen) provide aeronautical information about changes and deviations affecting airspace, aerodromes and air traffic. These include, among others, airspace changes, runway closures and radio frequency changes. The main objective of NOTAMs is to warn and protect all participants in aviation against possible hazards in order to maintain flight safety. NOTAMs consist of completely capitalized short free texts with many abbreviations. Dispatchers plan flights for the pilots having the goal to reduce their cognitive effort and improve flight safety. They have to take into account routes, aircraft type, weather, performance calculations as well as several pages of NOTAMs for flight planning. For each flight, they have to read, understand and process several hundred NOTAMs. Overlooking or misunderstanding of a NOTAM can have serious consequences for flight safety. Although emerging technologies continue to offer new opportunities, NOTAMs evolve slowly to not at all. Therefore, the aim of this work is to reduce the likelihood of overlooking important NOTAMs by facilitating the interaction with them. In order to achieve the best possible results for dispatchers, user-centered design methods were used. The requirements assessment and analysis then revealed the necessity of a prioritized NOTAM ranking. By classifying NOTAMs and assessing the importance of these classes with users, a graph structure could be created. It was then possible to apply PageRank [1], originally proposed for website ranking, to these graphs. This resulted in a ranking of NOTAMs prioritized by dispatchers. For each class an icon was created to visualize the rankings. In order to evaluate the ranking and the icons, a prototype was implemented. This prototype was then evaluated by an user experience expert and by dispatchers. The positive validation by the users indicates that a logical and meaningful prioritization of NOTAMs with easily understandable icons was found generated by a novel NOTAM ranking approach using PageRank.

Show publication details
Chandrashekara, Namitha; Kuijper, Arjan [1. Gutachten]; Burkhardt, Dirk [2. Gutachten]

User-Centered Scientific Publication Research and Exploration in Digital Libraries

2018

Darmstadt, TU, Master Thesis, 2018

Scientific research is the basis for innovations. Surveying the research papers is an essential step in the process of research. It is vital to elaborate the intended writing of state of the art. Due to the rapid growth in scientific and technical discoveries, there is an increasing availability of publications. The traditional method of publishing the research papers includes physical libraries and books. These become hard to document with the rise in the number of publications produced. Due to the above mentioned problem, online archives for scientific publications have become more prominent in the scientific community. The availability of the search engines and digital libraries help the researchers in identifying the scientific publications. However, they provide limited search capabilities and visual interface. Most of the search engines have a single field to search and provides basic filtering of the data. Therefore, even with popular search engines, it is hard for the user to survey the research papers as it limits the user to search based on simple keywords. The relationships across multiple fields of the publications are also not considered such as to find the related papers and papers based on the citations or references. The main aim of the thesis is to develop a visual access to the digital libraries based on the scientific research and exploration. It helps the user in writing scientific papers. A scientific research and exploration model is developed based on the previous information visualization model for visual trend analysis with digital libraries, and with consideration of the research process. The principles from Visual Seeking Mantra are incorporated to have an interactive user interface that enhances the user experience. In the scope of this work, a research on Human Computer Interaction, particularly considering the aspects of user interface design are done. An overview of the scientific research, its types and various aspects of data analysis are researched. Different research models, existing approaches and tools that help the researchers in literature survey are also researched. The architecture and the implementation details of scientific research and exploration that provides visual access to digital libraries are presented.

Show publication details
Korn, Oliver; Buchweitz, Lea; Rees, Adrian; Bieber, Gerald; Werner, Christian; Hauer, Klaus

Using Augmented Reality and Gamification to Empower Rehabilitation Activities and Elderly Persons. A Study Applying Design Thinking

2018

Advances in Artificial Intelligence, Software and Systems Engineering

International Conference on Applied Human Factors and Ergonomics (AHFE) <9, 2018, Orlando, Florida, USA>

We present the design of a system combining augmented reality (AR) and gamification to support elderly persons’ rehabilitation activities. The system is attached to the waist; it collects detailed movement data and at the same time augments the user’s path by projections. The projected AR-elements can provide location-based information or incite movement games. The collected data can be observed by therapists. Based on this data, the challenge level can be more frequently adapted, keeping up the patient’s motivation. The exercises can involve cognitive elements (for mild cognitive impairments), physiological elements (rehabilitation), or both. The overall vision is an individualized and gamified therapy. Thus, the system also offers application scenarios beyond re-habilitation in sports. In accordance with the methodology of design thinking, we present a first specification and a design vision based on inputs from business experts, gerontologists, physiologists, psychologists, game designers, cognitive scientists and computer scientists.

Show publication details
Zouhar, Florian; Kuijper, Arjan [Supervisor]; Senner, Ivo [Advisor]

Vector Based Web Visualization of Geospatial Big Data

2018

Darmstadt, TU, Bachelor Thesis, 2018

Today, big data is one of the most challenging topics in computer science. To give customers, developers or domain experts an overview of their data, one needs to visualize these. They need to explore their data, using visualization technologies on high level but also in detail. As base technology, visualizations can be used to do more complex data analytic tasks. In case data contains geospatial information it becomes more difficult, because nearly every user has a well trained experience how to explore geographic information. These map applications provide an interface, in which users can zoom and pan over the whole world. This thesis focuses on evaluating one approach to visualize huge sets of geospatial data in modern web browsers. The contribution of this work is, to make it possible to render over one million polygons integrated in a modern web application which is done by using 2D Vector Tiles. Another major challenge is the web application, which provides interaction features like data-driven filtering and styling of vector data for intuitive data exploration. The important point is memory management in modern web browsers and its limitations.

Show publication details
Blüml, Jannis Ralf Joachim; Kuijper, Arjan [1. Gutachten]; May, Thorsten [2. Gutachten]

Verbesserung von Visualisierungen aus Dimensionsreduktion mittels topographischer Strukturen

2018

Darmstadt, TU, Bachelor Thesis, 2018

Visualisierung und Dimensionsreduktion spielen haufig eine entscheidende Rolle bei der Verarbeitung von Daten durch den Menschen, haben aber in ihren aktuellen Formen einige Schwachstellen. Die meisten Dimensionsreduktionsverfahren stellen ihre Daten nur als Scatterplot dar. Die Nutzung von komplexeren, neueren Visualisierungstechniken ist eher unublich. So gehen allerdings Informationen uber die Daten verloren und konnen daher von Expert*innen nicht erkannt, genutzt werden. Weiter sind Reduktionen immer fehlerbehaftet, weshalb die Aussagekraft solcher Visualisierungen immer mit Vorsicht betrachtet werden muss. Im Rahmen dieser Arbeit wird ein Ansatz zur Darstellung groser Datenmengen untersucht, welcher Dimensionsreduktionsverfahren mit topographischer Visualisierung kombiniert. Ein wesentlicher Bestandteil ist dabei die Erstellung eines Prototyps, auf dessen Grundlage man den so erarbeiteten Ansatz mit herkommlichen Verfahren vergleichen kann. Fur den Vergleich werden bereits bekannte Metriken wie die Erhaltung von lokalen Nachbarschaften oder die Erhaltung von Distanzen verwendet. Fur das Verfahren selbst werden bereits vorhandene und etablierte und aktuelle Ansatze aus dem Bereich der Dimensionsreduktion eingesetzt. Die durch diesen Ansatz konstruierten Visualisierungen sollen eine Abbildung von Punkt-zu-Punkt Distanzen ermoglichen, welche adaquater ist bei herkommlichen Verfahren ublich. Fur die Darstellung verwenden wir topographische Strukturen, welche die Distanz zweier Punkte kunstlich erhohen. Die Distanz wird dabei nicht mehr durch die euklidische Distanz approximiert, sondern wie bei einer topographischen Karte gelesen. Es entsteht dadurch eine Hohenkarte der Daten. Die Nutzer*innen des Programmes sollen mit dessen Hilfe ein besseres Verstandnis der Daten erhalten und unglaubwurdigere Abbildungen erkennen konnen. Das Hauptaugenmerk dieser Arbeit liegt auf der Uberprufung der Machbarkeit eines solchen Ansatz sowie dem Testen und Evaluieren eines Prototyps.

Show publication details
Bernard, Jürgen; Zeppelzauer, Matthias; Sedlmair, Michael; Aigner, Wolfgang

VIAL: A Unified Process for Visual Interactive Labeling

2018

The Visual Computer

The assignment of labels to data instances is a fundamental prerequisite for many machine learning tasks. Moreover, labeling is a frequently applied process in visual interactive analysis approaches and visual analytics. However, the strategies for creating labels usually differ between these two fields. This raises the question whether synergies between the different approaches can be attained. In this paper, we study the process of labeling data instances with the user in the loop, from both the machine learning and visual interactive perspective. Based on a review of differences and commonalities, we propose the "visual interactive labeling" (VIAL) process that unifies both approaches.We describe the six major steps of the process and discuss their specific challenges. Additionally, we present two heterogeneous usage scenarios from the novel VIAL perspective, one on metric distance learning and one on object detection in videos. Finally, we discuss general challenges to VIAL and point out necessary work for the realization of future VIAL approaches.

Show publication details
Olbrich, Manuel; Graf, Holger; Keil, Jens; Gad, Rüdiger; Bamfaste, Steffen; Nicolini, Frank

Virtual Reality Based Space Operations – A Study of ESA’s Potential for VR Based Training and Simulation

2018

Virtual, Augmented and Mixed Reality: Interaction, Navigation, Visualization, Embodiment, and Simulation

International Conference Virtual Augmented and Mixed Reality (VAMR) <10, 2018, Las Vegas, NV, USA>

Lecture Notes in Computer Science (LNCS)
10909

This paper presents the results of a study the authors conducted together over a year in order to identify key issues of ESA’s (European Space Agency) potential for a deployment of Virtual Reality training environments within space operations. Typically, ESA simulates several operations using DES like systems that need to be linked to a VR environment for training purposes. Based on the second generation of VR equipment and development tools the paper describes a holistic design approach from scenario development through design decisions on SW and HW choices until the final development of a PoC for a virtual lunar base that might simulate the metabolism of a lunar base. Here the idea was to mirror the mass- and energy-flows within a lunar base in order to maintain an environment, in which astronauts can live and work and to establish a tool that supports the training of astronauts for operating such a lunar base, the one likely next step of human space exploration beyond the International Space Station as identified by ESAs decision makers. In the end, we have realized a PoC for a fire emergency case on a lunar base allowing astronauts being trained in a fully simulated and integrated environment. The system could be tested and evaluated in two set-ups, first using classical VR controllers, second, using recent VR glove technology.

  • 978-3-319-91580-7
Show publication details

Visual Analytics for Radiomics: Combining Medical Imaging with Patient Data for Clinical Research

2018

2017 IEEE Workshop on Visual Analytics in Healthcare

IEEE Workshop on Visual Analytics in Healthcare (VAHC) <8, 2017, Phoenix, USA>

The visualization and analysis of electronic health records (EHRs) are becoming increasingly relevant for clinical researchers. While the digitization of medical images is general practice today, many clinics are just starting to build up database with the related patient data, patient histories, and treatment outcomes. This paper reports on a project with a medical group of ear, nose, and throat (ENT) specialists. It combines medical image analysis and Radiomics with visual analytics of patient data to build, analyze, and evaluate patient cohorts. The combined visual interface for both browsing and analyzing patient data was developed in collaboration with the medical researchers. In addition to offering a new way of cohort building, our approach also provides a first comprehensive view on the EHR, including the relevant anatomy of patients. This project triggered a new effort to extend the digitized patient database from around 100 patients to the entire patient population at our partner’s clinic.

  • 978-1-5386-3187-4
Show publication details
Ruppert, Tobias; Fellner, Dieter W. [Referent]; Kohlhammer, Jörn [Referent]; Miksch, Silvia [Referentin]

Visual Analytics to Support Evidence-Based Decision Making

2018

Darmstadt, TU, Diss., 2018

The aim of this thesis is the design of visual analytics solutions to support evidence-based decision making. Due to the ever-growing complexity of the world, strategical decision making has become an increasingly challenging task. At the business level, decisions are not solely driven by economic factors anymore. Environmental and social aspects are also taken into account in modern business decisions. At the political level, sustainable decision making is additionally influenced by the public opinion, since politicians target the conservation of their power. Decision makers face the challenge of taking all these factors into consideration and, at the same time, of increasing their efficiency to immediately react on abrupt changes in their environment. Due to the digitization era, large amounts of data are digitally stored. The knowledge hidden in these datasets can be used to address the mentioned challenges in decision making. However, handling large datasets, extracting knowledge from them, and incorporating this knowledge into the decision making process poses significant challenges. Additional complexity is added by the varying expertises of stakeholders involved in the decision making process. Strategical decisions today are not solely made by individuals. In contrast, a consortium of advisers, domain experts, analysts, etc. support decision makers in their final choice. The amount of involved stakeholders bears the risk of hampering communication efficiency and effectiveness due to knowledge gaps coming from different expertise levels. Information systems research has reacted to these challenges by promoting research in computational decision support systems. However, recent research shows that most of the challenges remain unsolved. During the last decades, visual analytics has evolved as a research field for extracting knowledge from large datasets. Therefore, combining human perception capabilities and computers' processing power offers great analysis potential, also for decision making. However, despite obvious overlaps between decision making and visual analytics, theoretical foundations for applying visual analytics to decision making have been missing. In this thesis, we promote the augmentation of decision support systems with visual analytics. Our concept comprises a methodology for the design of visual analytics systems that target decision making support. Therefore, we first introduce a general decision making domain characterization, comprising the analysis of potential users, relevant data categories, and decision making tasks to be supported with visual analytics technologies. Second, we introduce a specialized design process for the development of visual analytics decision support systems. Third, we present two models on how visual analytics facilitates the bridging of knowledge gaps between stakeholders involved in the decision making process: one for decision making at the business level and one for political decision making. To prove the applicability of our concepts, we apply our design methodology in several design studies targeting concrete decision making support scenarios. The presented design studies cover the full range of data, user, and task categories characterized as relevant for decision making. Within these design studies, we first tailor our general decision making domain characterization to the specific domain problem at hand. We show that our concept supports a consistent characterization of user types, data categories and decision making tasks for specific scenarios. Second, each design study follows the design process presented in our concept. And third, the design studies demonstrate how to bridge knowledge gaps between stakeholders. The resulting visual analytics systems allow the incorporation of knowledge extracted from data into the decision making process and support the collaboration of stakeholders with varying levels of expertises.

Show publication details
Kohlhammer, Jörn; Proff, Dirk U.; Wiener, Andreas

Visual Business Analytics

2018

Edition TDWI

Business-Intelligence-Lösungen sind für Unternehmen unabdingbar, um Datenmengen in vertretbarer Zeit zu analysieren und daraus resultierend Entscheidungen zu treffen. Dieses Buch zeigt den Weg auf, wie aus Daten mittels Visualisierung entscheidungsrelevante Informationen für den Empfänger werden. Neue, interaktive und grafische Darstellungen tragen dazu bei, dass Entscheider ihr Wissen und ihre Fähigkeiten besser nutzen können, um einen echten Mehrwert für ihr Unternehmen zu generieren. Die Autoren bieten eine fundierte Einführung in das Thema und geben einen praxisnahen Überblick über Visual Business Analytics mit seinen drei Teilgebieten: Information Design, Visual Business Intelligence und Visual Analytics. Sie erläutern anhand vieler Beispiele aus Business-Intelligence-Anwendungsszenarien, welche Darstellungsformen jeweils geeignet sind, um komplexe Zusammenhänge abzubilden, wie Unternehmen Visual Business Analytics erfolgreich nutzen können und welche zukünftigen Möglichkeiten sich durch interaktive Darstellungen ergeben.

  • 978-3-86490-410-3
Show publication details
Brakowski, Alexander; Maier, Sebastian; Kuijper, Arjan

Visual Guidance to Find the Right Spot in Parameter Space

2018

Human Interface and the Management of Information. Interaction, Visualization, and Analytics

International Conference on Human Interface and the Management of Information (HIMI) <20, 2018, Las Vegas, NV, USA>

Lecture Notes in Computer Science (LNCS)
10904

The last few decades brought upon a technological revolution that has been generating data by users with an ever increasing variety of digital devices, resulting in such an incredible volume of data, that we are unable to make any sense of it any more. One solution to decrease the required execution time of these algorithms would be the preprocessing of the data by sampling it before starting the exploration process. That indeed does help, but one issue remains when using the available Machine Learning and Data Mining algorithms: they all have parameters. That is a big problem for most users, because a lot of these parameters require expert knowledge to be able to tune them. Even for expert users a lot of the parameter configurations highly depend on the data. In this work we will present a system that tackles that data exploration process from the angle of parameter space exploration. Here we use the active learning approach and iteratively try to query the user for their opinion of an algorithm execution. For that an end-user only has to express a preference for algorithm results presented to them in form of a visualisations. That way the system is iteratively learning the interest of the end-user, which results in good parameters at the end of the process. A good parametrisation is obviously very subjective here and only reflects the interest of an user. This solution has the nice ancillary property of omitting the requirement of expert knowledge when trying to explore an data set with Data Mining or Machine Learning algorithms. Optimally the end-user does not even know what kind of parameters the algorithms require.

Show publication details
Ballweg, Kathrin; Pohl, Margit; Wallner, Günter; Landesberger, Tatiana von

Visual Similarity Perception of Directed Acyclic Graphs: A Study on Infuencing Factors and Similarity Judgment Strategies

2018

Journal of Graph Algorithms and Applications

Visual comparison of directed acyclic graphs (DAGs) is commonly encountered in various disciplines (e.g., finance, biology). Still, knowledge about humans' perception of their similarity is currently quite limited. By similarity perception, we mean how humans perceive commonalities and differences of DAGs and herewith come to a similarity judgment. To fill this gap, we strive to identify factors influencing the DAG similarity perception. Therefore, we conducted a card sorting study employing a quantitative and qualitative analysis approach to identify (1) groups of DAGs the participants perceived as similar and (2) the reasons behind their groupings. We also did an extended analysis of our collected data to (1) reveal specifics of the influencing factors and (2) investigate which strategies are employed to come to a similarity judgment. Our results suggest that DAG similarity perception is mainly influenced by the number of levels, the number of nodes on a level, and the overall shape of the DAG. We also identified three strategies used by the participants to form groups of similar DAGs: divide and conquer, respecting the entire dataset and considering the factors one after the other, and considering a single factor. Factor specifics are, e.g., that humans on average consider four factors while judging the similarity of DAGs. Building an understanding of these processes may inform the design of comparative visualizations and strategies for interacting with them. The interaction strategies must allow the user to apply her similarity judgment strategy to the data. The considered factors bear information on, e.g., which factors are overlooked by humans and thus need to be highlighted by the visualization.

Show publication details
Purey, Ranveer; Kuijper, Arjan [1. Gutachten]; Burkhardt, Dirk [2. Gutachten]

Visual Trend Analysis on Digital Semantic Library Data for Innovation Management

2018

Darmstadt, TU, Master Thesis, 2018

The amount of scientific data published online has been witnessing massive growth in the recent years. This has lead to exponential growth in the amount of data stored in digital libraries(DLs) such as springer, eurographics, digital bibliography and library Project (dblp), etc. One of the major challenges is to prevent users from getting lost in irrelevant search results, when they try to retrieve information in order to get meaningful insights from these digital libraries. This problem is known as information overload. Other challenge is the quality of data in digital libraries. A quality of data can be affected by factors such as missing information, absence of links to external databases or data is not well structured, and the data is not semantically annotated. Apart from data quality, one more challenge is the fact that, there are tools available which help users in retrieving and visualizing the information from large data sets, but these tools lack one or the other basic requirements like data mining, visualizations, interaction techniques etc. These issues and challenges have lead to increase in the research in the field of visual analytics, it is a combination of data processing, information visualization, and human computer interaction disciplines. The main goal of this thesis is to overcome the information overload problem and the challenges mentioned above. This can be achieved by using digital library named SciGraph by springer, which serves as a very rich source of semantically annotated data. The data from SciGraph can be used in combination with data integration, data mining and information visualization techniques in order to aid users in decision making process and perform visual trend analysis on digital semantic library data. This concept would be designed and developed as a part of innovation management process, which helps transforming innovative ideas into reality using a structured process. In this thesis, a conceptual model for performing visual trend analysis on digital semantic library data as part of innovation management process had been proposed and implemented. In order to create the conceptual model, several disciplines such as human computer interaction, trend detection methods, user centered design, user experience and innovation management have been researched upon. In addition, evaluation of various information visualization tools for digital libraries has been carried out in order to find out and address the challenges faced by these tools. The conceptual model proposed in this thesis, combines the usage of semantic data with information visualization process and also follows structured innovation management process, in order to ensure that the concept and implementation (proof of concept) are valid, usable and valuable to the user.

Show publication details
Heilmann, Johannes; Kuijper, Arjan [1. Prüfer]; Wuest, Harald [2. Prüfer]

Visual-Inertial Model Target Tracking for Consumer Hardware

2018

Darmstadt, TU, Bachelor Thesis, 2018

This thesis explores visual-inertial tracking for the application in Augmented Reality. Combining vision based tracking with data from inertial sensors like accelerometer and gyroscope can result in a faster and more robust tracking system. The two types of data complement each other well. Tracking results in the form of 2D/3D correspondences from either a poster tracker or model tracker are combined with inertial data from the sensors of a Surface Pro 2 and fused in an Extended Kalman Filter. Different configurations are available. On the vision side there is a poster tracker using FAST and BRIEF for matching and the KLT for tracking. Alternatively a model tracker can be used. It builds a line model from a CAD model and tracks points on the lines in the image. On the inertial sensor side there are also two options. Either only gyroscope measurements can be used, or data from the gyroscope and an accelerometer can be included. The system is built to be used with consumer hardware. For this thesis a Microsoft Surface Pro 2 was used. This means that the data is not synchronised to a common clock and the inertial sensors are less accurate than those found in specialised hardware. The system is evaluated on a recorded image sequence and corresponding sensor data. The different parameters of the EKF models are tuned by experimentally minimising the RSME between EKF estimations and accurate baseline results. It is shown that combining visual and inertial data allows the vision system to rely on tracking less features. This results in reduced computational cost. But the inertial sensors of the Surface Pro 2 are not accurate enough to allow a camera pose estimation based on inertial data alone in case the camera tracking fails.

Show publication details

Visual-Interactive Identification of Anomalous IP-Block Behavior Using Geo-IP Data

2018

VizSec 2018

IEEE Symposium on Visualization for Cyber Security (VizSec) <15, 2018, Berlin, Germany>

Routing of network packets from one computer to another is the backbone of the internet and impacts the everyday life of many people. Although, this is a fully automated process it has many security issues. IP hijacks and misconfigurations occur very often and are difficult to detect. In the past visual analytics approaches aimed at detecting these phenomenons but only a few of these integrated geographical references. Geo-IP data is being used mostly as a lookup table which is an undervaluation of its capabilities. In this paper we present a visual-interactive system which only relies on Geo-IP data to create more awareness for this data source. We show that looking at Geo-IP data over time in combination with owner and location information of IP blocks already reveals suspicious cases. Together with our design study we also contribute a pre-processing algorithm for the Maxmind GeoIP2 City and ISP databases, to motivate the community to integrate this data source in future approaches.

  • 978-1-5386-8194-7
Show publication details
Krause, Silvio; Bieber, Gerald; Tominski, Christian; Kreuseler, Matthias

Visualisierung von digitalen, individuellen Gesundheitsdaten am Beispiel des Blutzuckerverlaufs bei gesunden Menschen

2018

WIWITA 2018. Proceedings

Wismarer Wirtschaftsinformatiktage (WIWITA) <11, 2018, Wismar, Germany>

Bei der Nahrungsaufnahme beeinflussen verschiedenste Nahrungsbestandteile die Konzentration des Blutzuckers. Durch die natürlichen Regelmechanismen, beispielsweise der körpereigenen Ausschüttung von Insulin, werden die Blutzuckerkonzentrationen auf einem gesunden Wert gehalten. Bei bestimmten Krankheitsbildern ist jedoch die Regulierung der Blutzuckerkonzentration gestört. Während bislang eine Blutzuckermessung nur durch Blutabnahme möglich war, so ist durch Einsatz von CGM Systemen (CGM - Continuous Glucose Monitoring Systems) eine kontinuierliche Blutzuckermessung möglich. Der Blutzuckerwert wird bei diesen Systemen durch eine Analyse der Zellflüssigkeit bestimmt. Die CGM Systeme werden derzeit für Diabetes mellitus Typ 1 Patienten eingesetzt, deren Blutzuckerspiegel durch das Fehlen der körpereigenen Insulinproduktion gestört ist. Auch bei Gesunden variiert der Blutzuckerspiegel vorwiegend durch Nahrungsaufnahme, Ernährungsstil und körperliche Aktivität. Diese Blutzuckeränderungen können digital abgespeichert werden und sind zu jeder Zeit bequem im Alltag messbar. Diese Daten können anschließend mit angepassten Methoden der Visualisierung repräsentiert werden. Die Messung des Blutzuckerspiegels bei gesunden Personen wird jedoch bislang noch nicht durchgeführt, obwohl die digitale Abbildung relevante Aussagen hinsichtlich des aktuellen und individuellen Stoffwechselverhaltens, Stress sowie der körperlichen Fitness liefern könnte. Anhand der Daten wäre auch eine Entwicklung von Präventivmaßnahmen möglich, wie z.B. die Vorbeugung von Diabetes. In dem Beitrag wird eine Anwendung vorgestellt, die digitale Gesundheitsdaten anhand von ausgewählten Visualisierungskonzepten darstellt, sowie deren potenzielle Möglichkeiten der Nutzung diskutiert.

Show publication details
Enders, Jan; Stork, André [1. Gutachten]; Redkin, Maxim [2. Gutachten]

Visualisierung von Subdivisionsvolumendaten

2018

Darmstadt, TU, Bachelor Thesis, 2018

Während Für Volumendaten in Form von Voxelgittern schon seit langem viele verschiedene Visualisierungsmethoden existieren, gilt das Gleiche leider nicht für andere Formen von Volumendaten.Eine solche alternative Form von Volumendaten sind Subdivisionsvolumen, die im Vergleich zu Voxeldaten gewisse Vorteile mit sich bringen, allen voran eine glatte Interpolation von Parametern entlang des Volumenmodells.In dieser Arbeit wird ein Verfahren für die Visualisierung von Subdivisionsvolumen vorgestellt, das für Direct Volume Rendering und für die Darstellung von Isoflächen geeignet ist.Dieses Verfahren basiert auf einer Diskretisierung des Subdivisionsvolumens in ein Voxelgitter, das dann mithilfe von Raycasting dargestellt wird.Es erreicht auf den gleichen Daten eine bessere Performanz als zwei vergleichbare Volumenrenderer.Das Verfahren zur Diskretisierung des Subdivisionsvolumens erzeugt allerdings eine gewisse Menge an Fehler, weil die glatte kubische Interpolation des Subdivisionsvolumenmodells nicht erhalten bleibt.

Show publication details

Visualization of Zoomable 2D Projections on the Web

2018

HCI in Business, Government, and Organizations

International Conference on Human-Computer Interaction in Business, Government and Organization (HCIBGO) <5, 2018, Las Vegas, NV, USA>

Lecture Notes in Computer Science (LNCS)
10923

The objective of the work is the research and development of a web-based visualization system for the creation and testing of zoomable projection cards. The basic idea is to project a multidimensional data set onto two dimensions using projection methods to represent it on a 2D surface. Based on the Card, Mackinlay, and Shneiderman visualization pipeline, a data processing model has been developed. For data processing various distance metrics, dimension reduction methods, zooming approaches as well as presentation concepts are considered. The peculiarities and considerations of the respective technology are discussed. A zooming approach allows large amounts of data to be displayed on a limited area. In order to better visualize connections within the data, concepts of presentation are discussed. The data points are represented as glyph-based objects or using color maps, various shapes, and sizes. Best practices about colormaps are discussed. In order to display large amounts of data in real time, a separation of the generation and visualization process takes place. During generation, a tabular file and selected configuration execute computationally-intensive transformation processes to create map material. Similar to Google Maps, the generated map material is represented by a visualization. Management concepts for managing various map sets as well as their generation and presentation are presented. A user interface can be used to create and visualize map material. The user uploads a tabular file into the system and chooses between different configuration parameters. Subsequently, this information is used to generate map material. The maps and various interaction options are provided in the visualization interface. Using various application examples, the advantages of this visualization system are presented.

Show publication details
Wambach, Johannes; Hergenröther, Elke [Referent]; Rapp, Stefan [Korreferent]; Wirtz, Andreas [Betreuer]

Vollautomatische Segmentierung der Zähne in Fernröntgenseitenbildern unter Verwendung statistischer Formmodelle

2018

Darmstadt, Hochschule, Master Thesis, 2018

Fernröntgenseitenbilder (FRS) sind in der Zahnmedizin und der Kieferorthopädie ein wichtiges Hilfsmittel für die Behandlungsplanung. Auf einem FRS wird der Kopf direkt von der Seite aufgenommen. Ein FRS wird unter anderem für die Analyse des Gesichtsschädelaufbaus, der Erfassung der vertikalen und sagittalen bzw. horizontalen Kieferlagebeziehungen und der dentalen Beziehungen verwendet. Für die Analyse eines FRS werden Punkte und Linien markiert und mit diesen Winkel und Strecken berechnet. Die manuelle Analyse von FRS-Aufnahmen ist sehr zeitaufwendig. Da es sich hierbei um die Erkennung von Bildmerkmalen auf 2D-Grauwertbildern handelt, ist es möglich die Analyse von FRS-Aufnahmen zu automatisieren. In dieser Arbeit wird ein System zur vollautomatischen Segmentierung der Zähne in Fernröntgenseitenbilder unter Verwendung statistischer Formmodelle vorgestellt. Aus manuell segmentierten FRS-Aufnahmen wird ein statistisches Modell mit artikuliertem Atlas trainiert. Es wird jeweils ein Modell für Oberkiefer und ein Modell für Unterkiefer gelernt. Es wird dann eine Initialisierung des mittleren Modells auf einem Testbild, anhand von erkannten Bildmerkmalen, durchgeführt. So wird die initiale Position, Rotation und Skalierung des mittleren Modells gefunden. Danach kann das mittlere Modell an ein Testbild angepasst werden. Die Anpassung verwendet die ermittelten Gradienten im Testbild und das gelernte Modellwissen. Die so erhaltenen Segmentierungen werden mit manuellen Segmentierungen verglichen. Der berechnete Mittelwert der Überlappung der automatischen und manuellen Segmentierungen des Ober- und Unterkiefers liegt bei knapp über 80 Prozent.

Show publication details
Schurig, Martin; Horsch, Thomas [Referent]; Braun, Michael [Korreferent]; Tausch, Reimar [Betreuer]

Vorkalibriertes Focus Stacking für schnelle teilautomatisierte 3D-Digitalisierung

2018

Darmstadt, Hochschule, Bachelor Thesis, 2018

In dieser Arbeit wird untersucht inwieweit Focus Stacking, die Schärfentiefeerweiterung bei Bildern, automatisierbar ist, sodass die so erhaltenen, mit ausreichender erweiterter Tiefe, scharfen Bilder zur photogrammetrischen Rekonstruktion eingesetzt werden können. Focus Stacking ist ein Verfahren bei dem Bilder, die aus demselben Blickwinkel aber mit unterschiedlichen Fokusdistanzen aufgenommen wurden, zu einem einzigen Bild kombiniert werden. Dieses kombinierte einzelne Bild vereint alle scharfen Bildbereiche der Eingabebilder. So entsteht ein Bild mit einer Tiefenschärfe die viel größer sein kann, als das bei der Aufnahme der Bilder möglich wäre. Dabei Teilt sich Focus Stacking in zwei Hauptschritte auf: das Ausrichten der Bilder zueinander und das anschließende Kombinieren der scharfen Bildbereiche. Zur Automatisierung wird ein Aufbau entwickelt, der für kleine Objekte (unter 10 cm) optimiert ist. Die Hauptbestandteile des Aufbaus sind die Fotokamera Canon 5DS R, ein Drehteller und eine Lichtquelle für möglichst homogene Ausleuchtung. Im Zuge der Automatisierung von Focus Stacking wird evaluiert inwiefern Daten vorab berechnet werden können, um den Prozess zu beschleunigen. Diese Vorkalibrierung kann auf zwei verschiedene Arten erfolgen: Die erste Methode entkoppelt die Berechnung, der für das Ausrichten der Bilder benötigten Parameter, von den eigentlich zu kombinierenden Bildern. Die zweite Methode abstrahiert das Ausrichten der Bilder weiter, indem die benötigten Parameter durch einen absoluten Wert, der das Verhalten des sich ändernden Fokus global beschriebt. Im Anschluss werden die so erhaltenen Bilder mit großer Schärfentiefe zur photogrammetrischen 3D-Rekonstruktion eingesetzt. Um das zu ermöglichen wird ein Kalibrierungsprozess zur Bestimmung intrinsischer Kameraparameter entwickelt. Diese Parameter werden zur notwendigen Entzerrung der Bilder eingesetzt. Die so erhaltenen 3D-Modelle werden untersucht und es werden Szenarien entwickelt bei denen der Einsatz von Focus Stacking eine Verbesserung bewirkt.

Show publication details
Brunner, Daniel; Kuijper, Arjan [1. Prüfer]; Ulmer, Alex [2. Prüfer]

Web-based Visual-Interactive Exploration of Network Data

2018

Darmstadt, TU, Bachelor Thesis, 2018

The amount of cyberattacks in Germany increased over the last years but many small and medium-sized enterprises can not afford Security Operation Centers (SOCs) to find and handle these attacks. In this work I created a visual interactive analysis tool as a web-application that is used to show different aspects of data with interactive visualizations to enable users to find irregular traffic and attacks. Since the amount of data can be enormous the main visualization shows the data in an aggregated form. To appeal to both experts and novices in network traffic I created two different single page interfaces with different visualizations and layouts that best suit their abilities. To find out if the interfaces fit their expected audiences best I conducted a user study where users had to complete tasks in both interfaces and tell which interface is better for experts or novices.

Show publication details
Gutbell, Ralf; Pandikow, Lars; Kuijper, Arjan

Web-Based Visualization Component for Geo-Information

2018

Human Interface and the Management of Information. Interaction, Visualization, and Analytics

International Conference on Human Interface and the Management of Information (HIMI) <20, 2018, Las Vegas, NV, USA>

Lecture Notes in Computer Science (LNCS)
10904

Three-dimensional visualization of maps is becoming an increasingly important issue on the Internet. The growing computing power of consumer devices and the establishment of new technologies like HTML5 and WebGL allow a plug-in free display of 3D geo applications directly in the browser. Existing software solutions like Google Earth or Cesium either lack the necessary customizability or fail to deliver a realistic representation of the world. In this work a browser-based visualization component for geo-information is designed and a prototype is implemented in the gaming engine Unity3D. Unity3D allows translating the implementation to JavaScript and to embed it in the browser with WebGL. A comparison of the prototype with the opensource geo-visualization framework Cesium shows, that while maintaining an acceptable performance an improvement of the visual quality is achieved. Another reason to use a gaming engine as platform for our streaming algorithm is that they usually feature engines for physics, audio, traffic simulations and more, which we want to use in our future work.

  • 978-3-319-92042-9
Show publication details

What Can a Single Minutia Tell about Gender?

2018

2018 International Workshop on Biometrics and Forensics (IWBF)

International Workshop on Biometrics and Forensics (IWBF) <2018, Sassari, Italy>

Since fingerprints are one of the most widely deployed biometrics, several applications can benefit from an accurate fingerprint gender estimation. Previous work mainly tackled the task of gender estimation based on complete fingerprints. However, partial fingerprint captures are frequently occurring in many applications including forensics and consumer electronics, with the considered ratio of the fingerprint is variable. Therefore, this work investigates gender estimation on a small, detectable, and well-defined partition of a fingerprint. It investigates gender estimation on the level of a single minutia. Working on this level, we propose a feature extraction process that is able to deal with the rotation and translation invariance problems of fingerprints. This is evaluated on a publicly available database and with five different binary classifiers. As a result, the information of a single minutia achieves a comparable accuracy on the gender classification task as previous work using quarters of aligned fingerprints with an average of more than 25 minutiae.