Selected work

BarPath AI

Tracking barbell trajectories from training footage, with annotated video exports for review.

PythonOpenCVStreamlitVideo processing
Squat tracking demonstration published in the BarPath AI repositoryStill from the recorded project demo

From input to output.

  1. 01Training footage
  2. 02Barbell detection
  3. 03Trajectory smoothing
  4. 04Annotated export

What can the path of a barbell tell us about a lift?

The approach

Video frames pass through an OpenCV detector and centroid smoothing before the app draws the bar path. The repository also documents repetition counting, pixel-based velocity, and an optional adapter for custom YOLO weights.

Scope & limitations

The default detector is an OpenCV baseline. Velocity is measured in pixels per second; calibrated physical velocity is listed as future work in the repository.

SIGNALWATCH