Built from annotated images representing six Brazilian Jiu-Jitsu positions: guard, mount, side control, turtle, takedown, and standing. Each image includes category labels, bounding boxes, and keypoints formatted according to the COCO standard across training, validation, and test splits.
Use Cases
- Train a pose estimation model to identify joint locations in grappling scenarios using the keypoints data
- Develop an object detection system to classify BJJ positions using the category labels and bounding boxes
- Evaluate the performance of action recognition algorithms on ground-based human interactions using the test split
Strengths
- Includes annotations for 6 distinct grappling positions: guard, mount, side control, turtle, takedown, and standing
- Data is provided in COCO-style format including category labels, bounding boxes, and keypoints
- Features a pre-defined split of 80% training, 10% validation, and 10% testing, balanced per-category