Broiler-Net

In this project, we developed Broiler-Net, an AI-driven framework for real-time analysis of broiler chicken behavior in cage-free poultry houses. Our system integrates a deep convolutional object detector with a lightweight centroid-based tracking algorithm to continuously monitor individual chickens across video frames. By analyzing spatial proximity and motion patterns, we automatically detect two critical abnormal behaviors: huddling, which can indicate poor environmental conditions and increased mortality risk, and inactivity, an early marker of illness. The framework is optimized for efficiency, enabling deployment on edge devices in real farm environments. Our experimental results demonstrate high detection accuracy, highlighting the potential of AI-based monitoring systems to improve animal welfare, reduce economic losses, and support data-driven farm management.

Our team participated in the scientific–industrial competition at the 12th International Conference on Computer and Knowledge Engineering (ICCKE 2022). Our project was recognized for its technical quality and real-world applicability, and we were awarded 2nd place among the participating teams. The competition was sponsored by Pocare Company and focused on bridging academic research with industrial impact.