SAM-3D-Body Auto-Labeling for Person Detection
An auto-labeling pipeline that uses SAM-3D-Body outputs to generate person detection annotations for tennis-court scenes.
Project Snapshot
| Signal | Details |
|---|---|
| Timeline | Apr 2026 to May 2026 |
| Organization | Enhanced Robotics / Tenniix |
| Focus | SAM-3D-Body-Based Auto-Labeling Pipeline for Person Detection Dataset Generation |
| Stack | SAM-3D-Body |
Filtering Logic
| Rule | Purpose |
|---|---|
| Keep detections between 5 m and 40 m | Preserve people expected between the net and baseline |
| Remove detections outside the valid court range | Reduce false positives and poor-quality labels |
| Train on cleaned annotations | Improve the final person detector |
Impact
The pipeline converts SAM-3D-Body predictions into cleaner person-detection labels with distance-based filtering tailored to the tennis-court setup.