YORU: Animal behavior detection with object-based approach for real-time closed-loop feedback

Area of Science:

• Neuroethology
• Animal Behavior Analysis
• Deep Learning

Background:

• Analyzing complex social behaviors in animals is challenging.
• Conventional methods struggle with dynamic, multi-individual interactions.
• Deep learning offers potential for advanced behavior analysis.

Purpose of the Study:

• Introduce YORU (your optimal recognition utility), a novel deep learning approach for animal behavior detection.
• Enable robust and accurate classification of social behaviors across diverse species.
• Facilitate real-time behavior analysis and closed-loop feedback systems.

Main Methods:

• Utilized an object detection deep learning algorithm.
• Developed YORU to identify behaviors as "behavior objects" based on animal shape.
• Applied the method to classify social behaviors in vertebrates and insects.

Main Results:

• YORU successfully classified various social behaviors across species.
• Achieved real-time behavior analysis and closed-loop feedback capabilities.
• Demonstrated real-time photostimulation feedback delivery during close-proximity social interactions.

Conclusions:

• YORU provides a robust alternative to conventional pose estimation methods.
• The system overcomes challenges in analyzing multi-individual animal social behaviors.
• YORU advances neuroethology by enabling precise, real-time behavioral analysis and intervention.