Programming and Setting Up the Object Detection Algorithm YOLO to Determine Feeding Activities of Beef Cattle: A...

Area of Science:

• Animal Science
• Computer Vision
• Machine Learning

Background:

• Accurate monitoring of cattle feeding behavior is crucial for livestock management and welfare.
• Object detection algorithms offer a promising non-invasive method for behavior analysis.
• Previous studies have explored various computer vision techniques for animal behavior monitoring.

Purpose of the Study:

• To compare the performance of YOLOv8 and YOLOv10 object detection algorithms for identifying cattle feeding behaviors.
• To evaluate the effectiveness of these algorithms in a real-world farm setting.
• To determine the most suitable YOLO version for practical cattle behavior monitoring.

Main Methods:

• Videos of six Charolais bulls were recorded on a French farm.
• Three feeding behaviors (biting, chewing, visiting) were identified and labeled using Roboflow.
• Object detection performance was evaluated using YOLOv8 and YOLOv10, comparing precision, recall, and mAP scores.

Main Results:

• YOLOv10 demonstrated slightly higher precision, recall, mAP50, and mAP50-95 scores compared to YOLOv8.
• Both algorithms achieved an overall accuracy of approximately 90%.
• YOLOv8 trained faster and showed less overfitting, while YOLOv10 exhibited better consistency in predictions.

Conclusions:

• Both YOLOv8 and YOLOv10 are effective for detecting cattle feeding behaviors.
• YOLOv10 shows superior average performance, learning rate, and speed, making it more suitable for field applications.
• Further research could explore real-time implementation and larger datasets for enhanced accuracy.