Golf Swing Analysis Moves from Video Review to Real-Time Skeletal Tracking
Bryson DeChambeau partnered with Google Cloud to deploy a deep learning system that uses proprietary 2D and 3D models to track more than 30 body, club, and ball landmarks per frame. The model labels top of swing, impact, follow-through, and finish, then outputs outcome metrics for each shot. The full pipeline runs on Google Cloud infrastructure and feeds directly into coaching dashboards.
Athletes and coaches stop guessing from slow-motion footage and start iterating on quantified joint angles and club paths within seconds. Workflows shift from post-session film review to live, data-driven adjustments that compress the feedback loop from hours to minutes.
Bryson DeChambeau and the Google Cloud sports performance team have published swing telemetry that shows measurable reductions in dispersion on tracked shots during practice rounds.
Step 1: Open Google Cloud Console and enable the Media Translation and Vision AI APIs. Step 2: Upload a 1080p swing video to a Cloud Storage bucket and run the Pose Detection model to extract 33 skeletal keypoints plus club and ball markers. Step 3: Export the JSON output to BigQuery, then query average shoulder rotation at impact; the dashboard displays the exact degree change needed for the next swing. URL: https://cloud.google.com/vision/docs/people-poses