|
|
 |
| |
|
|
|
Neural network methods for the detection of farm animals in dense dynamic groups on images |
|
|
|
PP: 241-249 |
|
|
|
doi:10.18576/amis/180204
|
|
|
|
Author(s) |
|
|
|
Olga Ivashchuk,
Lyazzat Atymtayeva,
Alexei Zhigalov,
Bagdat Yagalieva,
Oleg Ivashchuk,
Vyacheslav Fedorov,
|
|
|
|
Abstract |
|
|
| Currently, there is an urgent need for non-invasive monitoring of farm animals’ health status, enabling swift responses to adverse situations such as morbidity, feeding disorders, and aggression. Globally, technologies for video monitoring of animals are being developed, which include image processing using intelligent methods, especially artificial neural networks. This paper presents the results of developing and investigating methods and models for detecting (individually identifying) farm animals, with a focus on pigs as a case study. These animals are located in dense, dynamic groups within agricultural complexes where traditional identification methods are less effective. To overcome this challenge, advanced neural network architectures, specifically Faster R-CNN and YOLOv5, were selected, finely tuned, and trained. The application of the YOLOv5 network achieved a detection accuracy with a mean Average Precision (mAP) of 94.05%, surpassing the accuracy demonstrated in comparable studies. These results provide a foundation for a hardware-software complex designed for non-invasive, automated monitoring of animal conditions, integrating intelligent data analysis. This system offers crucial support for science-based decision-making in the fields of animal husbandry and food security management.
|
|
|
|
|
|
|
|
