Deep-learning based identification, tracking, pose estimation, and behavior classification of interacting primates...

Area of Science:

• Neuroscience
• Ethology
• Computer Science

Background:

• Quantifying animal behavior from video is crucial for studying brain function, pharmacology, and genetics.
• Current methods struggle with analyzing group behaviors in complex environments.

Purpose of the Study:

• To introduce a novel deep learning architecture for classifying individual and social animal behavior directly from raw video frames.
• To develop a comprehensive pipeline (SIPEC) for automated behavioral analysis in complex settings.

Main Methods:

• A deep learning pipeline (SIPEC) was created, integrating segmentation, identification, pose-estimation, and classification.
• The system analyzes raw video frames, requiring minimal human supervision after initial training.
• It utilizes data from simple mono-vision cameras in home-cage setups.

Main Results:

• SIPEC accurately classifies individual and social behaviors of freely moving mice and socially interacting non-human primates in 3D.
• The system outperforms existing state-of-the-art methods in complex environments.
• No intervention is needed post-initial human supervision.

Conclusions:

• SIPEC offers a powerful, automated solution for complex animal behavior quantification.
• This approach advances research in neuroscience, pharmacology, and genetics by enabling detailed behavioral analysis.
• The system's effectiveness in complex environments and with simple cameras broadens its applicability.