← Back to Projects Surfing Pose Tracking & Maneuver Recognition
Analyze surfing videos with robust pose tracking and temporal maneuver recognition for coaching-style feedback.
Categories
CVML
Tech Used
MediaPipe PoseopenPoseRTMOYOLOv8-Posedeep learningclassification modelsPyTorchTensorFlowOpenCVONNX RuntimeTensorRTCUDANumPy/SciPyFastAPIAWSDockerGitHubpoetry
Problem
Surf footage is difficult to analyze because of water spray, camera motion, occlusions, and fast athlete movement, yet coaches still need reliable feedback signals.
Approach
- Implemented multiple pose-estimation backends to improve keypoint tracking under difficult footage
- Smoothed temporal trajectories and engineered motion features from joint geometry and movement cues
- Built a sequence-modeling pipeline for maneuver recognition and real-time feedback delivery
Results
- More stable pose-based analysis in challenging sports video
- Product-oriented foundation for maneuver detection and coaching feedback
Demo Videos