Research Select Project

DeepLabCut Imaging Pipeline

Machine learning-based pose estimation system for tracking zebrafish behavior in custom experimental rigs, enabling high-throughput behavioral analysis.

Source ↗
Python DeepLabCut Computer Vision Pose Estimation

Built a custom ML-based pose estimation system using DeepLabCut to track zebrafish behavior in a purpose-built experimental rig.

Methodology

  • DeepLabCut neural network for markerless pose estimation
  • Custom training pipeline on labeled video frames
  • Automated extraction of behavioral metrics from tracked keypoints
  • High-throughput analysis of locomotion patterns and stimulus responses

Impact

  • Enabled quantitative behavioral phenotyping previously done by manual observation
  • Processed thousands of frames per experiment with consistent accuracy
  • Reproducible tracking across multiple experimental sessions