Research and Application of YOLOv10 Algorithm Based on Image Recognition
Mengyuan Wang, Zejian Liang, Hong Huang, Aimin Liang, Hongyu Sun, Yunxiang Zhao · 2024
The aim of this study is to explore the application of deep learning techniques in the field of wildlife image recognition, especially by comparing and analyzing the performance of three models, YOLOv5, YOLOv8, and YOLOv10, with a view to finding the most suitable deep learning model for the monitoring of Class I protected animals. In our study, we constructed a deep learning model based on the YOLO family, and the YOLO algorithm significantly improves the efficiency and accuracy of detection by treating the target detection task as a one-shot regression problem and mapping directly from image pixels to bounding box coordinates and category probabilities. Our study focuses on the YOLOv10 model, which demonstrates superiority over the precision, recall, and mAP metrics than the YOLOv5 and YOLOv8 in terms of precision, recall and mAP metrics, and especially shows better robustness when dealing with datasets with unbalanced categories. The experimental results show that YOLOv10 achieves high accuracy in the task of class I protected animal image recognition, making it a powerful tool for protected animal class recognition. Through these studies, we not only provide a scientific basis for conservation work, but also open up a new path for the application of image recognition technology in the field of animal conservation, contributing to global biodiversity conservation.