Automated 3D Kinematic Monitoring for Circadian Activity and Anomaly Detection in Juvenile Fish

The article presents a high-throughput 3D behavioral phenotyping framework for monitoring juvenile tilapia, integrating deep learning object detection with binocular stereo vision. This system automates body length estimation and reconstructs 3D swimming trajectories, enabling precise quantification of 3D velocity and acceleration, which is crucial for establishing circadian locomotor baselines and detecting physiological stress. This advancement addresses the phenotyping bottleneck in precision aquaculture, offering a novel method for assessing fish vitality in high-density environments.