Animal group behavior analysis and pose estimation based on deep learning

Animal behavior analysis plays a pivotal role in neuroscience, behavioral ecology, animal welfare, and precision agriculture. However, traditional manual observation methods are often subjective, labor-intensive, and insufficient for large-scale quantification. The advent of deep learning has revolutionized this field, enabling automated, high-throughput and accurate analysis particularly in complex group settings. This review provides a comprehensive overview of recent advances in deep learning-based animal group pose estimation and behavior analysis. It systematically outlines the key stages from data acquisition to behavior interpretation, including object detection, multi-animal tracking, pose estimation, and individual identification. Representative models and tools are critically evaluated, along with their applications across various species and experimental contexts. While notable advancements have been attainedincluding refined occlusion handling via part affinity fields and augmented temporal behavior recognition through video transformersseveral core challenges persist. These include robustness in wild environments, rare behavior detection and long-term identity preservation. Future research should focus on end-to-end joint modeling, data-efficient learning paradigms and multimodal data integration for advancing robust and intelligent systems. This review aims to provide researchers with a panoramic view of the field, highlighting key methodologies and directions for future development.