Research progress of animal behavior recognition algorithms based on deep learning
Xianzhou Zeng, Mingxiong Gong, Yonghao Yin, Yiting Zhao, Songsheng Zhu · Digital Medicine · 2025
The recognition of animal behavior can enhance the accuracy and repeatability of animal experiments, provide reliable data and conclusions for scientific research, which has significant theoretical research implications and practical application value. Compared with human behavior analysis, animal behavior analysis develops slowly due to the limitation of labeled datasets, and research is often restricted to target detection. However, in recent years, the research of skeleton modeling based on deep learning can provide skeleton information and motion characteristics for animal behavior analysis, which has become an important method in the field of animal behavior research. This article integrates the latest research findings on animal behavior recognition using deep learning algorithms, with a focus on typical algorithms for pose estimation and skeleton modeling. Then, the typical algorithms and general evaluation indicators are sorted out, and their performance is compared on common datasets. Finally, the challenging and pressing issues in this field are summarized, and the future development trend is prospected.