Zobrazit v EDS

The group-housed pigs attacking and daily behaviors detection and tracking based on improved YOLOv5s and DeepSORT.

Uloženo v:
Podrobná bibliografie
Název: The group-housed pigs attacking and daily behaviors detection and tracking based on improved YOLOv5s and DeepSORT.
Autoři: Cheng, Tianyu, Sun, Fujie, Mao, Liang, Ou, Haoxuan, Tu, Shuqin, Yuan, Fang, Yang, Hairan
Zdroj: PLoS ONE; 10/31/2025, Vol. 20 Issue 10, p1-17, 17p
Témata: SWINE, ANIMAL behavior, AGRICULTURAL technology, VIDEO surveillance, OBJECT tracking (Computer vision), OBJECT recognition (Computer vision), DEEP learning
Abstrakt: Automatic detection and tracking of pig behaviors through video surveillance remain challenges due to farm demanding conditions, e.g., illumination conditions and occlusion of one pig from another. The main goal of this study is to develop a deep learning method based on the improved YOLOv5s and DeepSORT to detect and track the behaviors of pigs, which has the advantages of stability and high accuracy. Firstly, YOLOv5s with the attention mechanism is used for pig detection and behavior recognition. To deal with the missed detection and false detection due to occlusion and overlapping between pigs and pigs, the improved YOLOv5s adopts the Shape-IoU to optimize the bounding box regression loss function, which improves the robustness of the model. Then, the improved DeepSORT model is proposed to track each pig behaviors including eat, stand, lie and attack four behavior types. Finally, we conduct a comparison test under different lighting and density conditions for pig detection and behavior tracking on special dataset. Experimental results show that the mAP@0.5% of improved YOLOv5s algorithm increases from 92.7% to 99.3%, which means 6.6% accuracy improvement compared with the YOLOv5s model. In terms of tracking, the values of MOTA and MOTP in all test videos are 94.5% and 94.9% respectively. These experiments demonstrate that the improved YOLOv5s and DeepSORT achieves high accuracy for both pig detection and behavior tracking. The proposed approach provides scalable technical support for contactless automatic pig monitoring. [ABSTRACT FROM AUTHOR]
Copyright of PLoS ONE is the property of Public Library of Science and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Databáze: Complementary Index
Popis
Abstrakt:Automatic detection and tracking of pig behaviors through video surveillance remain challenges due to farm demanding conditions, e.g., illumination conditions and occlusion of one pig from another. The main goal of this study is to develop a deep learning method based on the improved YOLOv5s and DeepSORT to detect and track the behaviors of pigs, which has the advantages of stability and high accuracy. Firstly, YOLOv5s with the attention mechanism is used for pig detection and behavior recognition. To deal with the missed detection and false detection due to occlusion and overlapping between pigs and pigs, the improved YOLOv5s adopts the Shape-IoU to optimize the bounding box regression loss function, which improves the robustness of the model. Then, the improved DeepSORT model is proposed to track each pig behaviors including eat, stand, lie and attack four behavior types. Finally, we conduct a comparison test under different lighting and density conditions for pig detection and behavior tracking on special dataset. Experimental results show that the mAP@0.5% of improved YOLOv5s algorithm increases from 92.7% to 99.3%, which means 6.6% accuracy improvement compared with the YOLOv5s model. In terms of tracking, the values of MOTA and MOTP in all test videos are 94.5% and 94.9% respectively. These experiments demonstrate that the improved YOLOv5s and DeepSORT achieves high accuracy for both pig detection and behavior tracking. The proposed approach provides scalable technical support for contactless automatic pig monitoring. [ABSTRACT FROM AUTHOR]
ISSN:19326203
DOI:10.1371/journal.pone.0334783