Mislaying behavior detection in cage-free hens with deep learning technologies.

Area of Science:

• Animal Science
• Agricultural Engineering
• Computer Science

Background:

• Floor egg-laying behavior (FELB) is a significant issue in commercial cage-free (CF) poultry houses, leading to increased labor and food safety concerns.
• Poor management practices can result in up to 10% of daily eggs being laid on the floor, necessitating improved detection and management strategies.

Purpose of the Study:

• To develop and evaluate a novel deep-learning model for the timely detection of FELB in research CF houses.
• To compare the performance of various YOLOv5 models in identifying floor eggs and assess factors influencing detection accuracy.

Main Methods:

• Five YOLOv5 deep-learning models (YOLOv5n, YOLOv5s, YOLOv5m, YOLOv5l, YOLOv5x) were trained and tested using a dataset of 5400 images from research CF houses.
• Model performance was evaluated based on precision, recall, mAP@0.50, and F1-score, with additional analysis of processing speed, training time, and GPU requirements.
• The impact of camera height (0.5 m vs. 3 m) and camera condition (clean vs. dusty) on detection was also assessed.

Main Results:

• YOLOv5m-FELB and YOLOv5x-FELB models demonstrated high performance with 99.9% precision and 99.2% recall.
• The YOLOv5s model offered faster data processing (4%-45% FPS) and shorter training times (3%-148%) with lower GPU usage, making it a promising candidate for further innovation.
• Optimal detection was achieved with a camera height of 0.5 m and a clean camera, outperforming higher or dustier camera setups.

Conclusions:

• Deep learning models, particularly YOLOv5 variants, show significant potential for accurately detecting floor egg-laying behavior in cage-free poultry systems.
• The YOLOv5s model presents a balance of detection accuracy and computational efficiency, suitable for further development and potential commercial application.
• Future research should focus on validating the model's performance in commercial settings to mitigate economic losses and enhance food safety in the poultry industry.