Tracker.
0 JOBS
No tracking jobs yet
Upload a video clip and select tracking methods — GVHMR for 3D
body mesh, DWPose for skeleton, or depth point cloud.
AI motion tracking pipeline — 3D body mesh, hand tracking, pose estimation, and depth reconstruction from monocular video.
World-Grounded Human Mesh Recovery (SIGGRAPH Asia 2024). Recovers full 3D SMPL body meshes with world-space positioning and ground plane attachment. Includes automatic WiLoR 3D hand tracking (MANO finger pose, attached to the body), Savitzky-Golay temporal smoothing, and foot-contact ground pinning.
DepthAnything2 monocular depth estimation, masked by a SAM3 subject matte and back-projected into a 3D point cloud (.ply) of just the isolated subject — the rest of the scene is discarded (depth + matte fusion). Includes a grayscale depth preview video. Useful for volumetric subject capture and spatial analysis.
YOLOv8-Pose for real-time 2D skeleton extraction. Detects 17 body keypoints (COCO format) per person per frame with confidence scoring. Includes temporal smoothing (Savitzky-Golay) to reduce jitter between frames.
Full COCO-WholeBody keypoint detection: 17 body + 68 face + 42 hand (21 per hand) + 6 foot keypoints. Uses the DWPose model for comprehensive whole-body motion capture including finger articulation and facial landmark tracking.
WiLoR 3D hand mesh recovery. Per-hand MANO pose (15 finger joints, axis-angle) recovered from a single image, robust on in-the-wild hands. When a body track (GVHMR) is present the hand pose merges into the SMPL-X body; otherwise it runs standalone. Tracks both hands.
Solves the shot camera from the plate — locked-off (static) or moving (MegaSaM SLAM + metric depth). Auto-detect picks the mode from measured camera motion. Focal length is estimated automatically (UniDepth) — no sensor or lens input needed.
Select one or more methods per job. Each runs independently, routed to a free GPU (Thor or linuxdev):