Deep Learning Maps 30-Plus Points to Perfect the Swing
DeChambeau's AI system uses deep learning with proprietary 2D and 3D models to track over 30 keypoints on the body, club, and ball. It records swing phases including top of swing, impact, follow-through, and finish, then logs ball-flight outcomes. All processing runs on Google Cloud infrastructure.
Users see that temporal segmentation of a motion sequence converts raw video into labeled training data. This changes practice from repetition without feedback to repetition with instant phase-specific metrics. Analysts gain a repeatable pipeline for turning smartphone footage into structured performance datasets.
Bryson DeChambeau's performance team runs the pipeline on Google Cloud. The setup has produced quantifiable improvements in club-path consistency during tournament weeks.
Step 1: Record a swing video on any smartphone and upload it to a Google Cloud Storage bucket. Step 2: Invoke the Sportsbox inference endpoint via the Google Cloud console or API; the model returns JSON with 30-plus keypoints and phase labels. Step 3: Parse the JSON in a Colab notebook to plot club path over time and compare against target values at cloud.google.com.