Last released Aug 9, 2026
Multi-task vision model: boxes, polygons, keypoints, polylines, and image tags in one network
Supported by