Poultry Detection on Hooks in Meat Processing Plant Production Lines

This study focuses on automating chicken production monitoring in a meatpacking plant, applying YOLO to detect birds on hooks on production lines. We detect two objects in authentic images: a chicken (birds hanging on the hook) and a hook (the metal support structure). By counting these objects in videos, we can monitor daily production and identify faults on the production line. We annotated 400 images for the experiment and used 320 to train seven YOLO versions, including lightweight variants (nano, tiny, and small) optimized for low-end devices. In testing, YOLOv9-tiny achieved 0.9967 precision but had lower recall than other experimental instances. YOLOv11-nano achieved an F1-score of 0.9924 and a mAP@50 of 0.9947, standing out in detection quality. Despite minor differences, the results were close for the YOLO versions analyzed, demonstrating a considerable ability to predict the analyzed objects.