DeepRod: A human-in-the-loop system for automatic rodent behavior analysis

This study presents a UX-optimized platform for automated rodent behavior analysis in preclinical drug discovery. The platform integrates AI-based behavior prediction, active learning for rare event detection, and novel behavior recognition, addressing the impracticality of manually analyzing extensive video data. Utilizing a cloud-native processing pipeline and a two-stage machine learning model, the system significantly improves annotation efficiency and model performance, advancing the detection and classification of complex rodent behaviors.