Automatic Detection of Cage-Free Dead Hens with Deep Learning Methods

Poultry farming plays a significant role in ensuring food security and economic growth many countries. However, various factors such as feeding management practices, environmental conditions, diseases lead to poultry mortality (dead birds). Therefore, regular monitoring of flocks timely veterinary assistance is crucial for maintaining health, well-being, the success operations. current method relies on manual inspection by farm workers, which time-consuming. developing an automatic early detection (MD) model with higher accuracy necessary prevent spread infectious poultry. This study aimed develop, evaluate, test performance YOLOv5-MD YOLOv6-MD models detecting under cage-free (CF) housing settings, including camera height, litter condition, feather coverage. The results demonstrated that YOLOv5s-MD performed exceptionally well, achieving high [email protected] score 99.5%, FPS 55.6, low GPU usage 1.04 GB, fast-processing time 0.4 h. Furthermore, this also evaluated models’ performances different CF levels coverage, height. 0% feathered covering achieved best overall object detection, highest 99.4% precision rate 98.4%. 80% resulted MD. Additionally, 100% recall hens’ at height 0.5 m but faced challenges greater heights 2 m. These findings suggest can detect more accurately than other models, its be optimized adjusting settings. developed assist farmers promptly responding events isolating affected birds, implementing disease prevention measures, seeking assistance, thereby helping reduce impact industry, well-being