Computer Vision–Based Sex Classification in Broiler Breeder Chickens

Initial Situation and Objectives

In broiler breeder management, sex-specific flock management is of central importance, particularly for management-relevant measures such as needs-based feeding, evaluating flock development, and assessing animal welfare and performance.

Currently, key flock parameters such as biomass, body condition, or population-based performance indicators are collected with a high level of manual effort. In many cases, male and female animals are considered together, even though they differ significantly in their physical development. The lack of sex-specific differentiation leads to distorted population-level indicators and makes targeted management more difficult.

The goal of the project is to develop a robust, camera-based system for the automated sex classification of broiler breeder chickens, enabling precise and practical separation of male and female animals as a basis for differentiated flock management.

Technological Approach

The project develops computer vision and machine learning methods for the automated classification of broiler breeder chickens by sex (male/female). The system is based on image data collected in poultry houses under real operational conditions.

The system uses:

• RGB and night-vision images for detection independent of lighting conditions
• AI models for robust differentiation between male and female animals
• Methods for stable classification under dynamic lighting conditions and varying environmental factors (temperature, humidity)

The solution is designed for broiler breeder chickens from approximately 18 weeks of age and is intended to operate autonomously without requiring a permanent internet connection. The target accuracy is comparable to that of experienced professionals.

Benefits, Application, and Exploitation

Automated sex classification enables sex-specific derivation of key flock parameters, including:

• population-based biomass and body condition indicators
• sex-specific behavioral analyses
• a differentiated assessment of flock development

Based on these insights, management measures—such as feeding strategies or flock evaluation—can be implemented more precisely and in a more targeted manner. The results are evaluated in aggregated form and provided as sex-specific distribution information, thereby improving the decision-making basis for flock management.

The solution is designed for seamless system integration and can be incorporated as a software module into existing mobile or stationary management and monitoring systems. In the long term, exploitation is planned through integration into commercial solutions offered by system providers.

Practical Environment

The development is carried out in collaboration with Big Dutchman International GmbH, an industrial partner with extensive expertise in poultry farming and housing technology.