zju3dv/EasyMocap

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

4783 stars · 560 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

EasyMocap is an open-source Python toolbox for markerless human motion capture and novel view synthesis from RGB videos. It fits parametric body models like SMPL, SMPL-X, and MANO using single-view internet videos or calibrated multi-view camera setups.

Use cases

• capture 3d human motion from multiple synchronized rgb cameras
• fit smpl or smpl-x body models to 2d keypoints from video
• reconstruct body hand and face poses without markers or suits
• estimate human motion from a single internet video
• recover 3d pose from a video of a person in front of a mirror
• generate novel views of a human from sparse multi-view videos
• build a motion capture dataset for animation or research

When to choose

• you need markerless motion capture from rgb video without wearable sensors
• you have calibrated multi-view cameras and want high-quality smpl fits
• you want to extract 3d human pose from ordinary youtube or single-camera footage
• you are doing computer vision research on human body reconstruction or novel view synthesis

When to avoid

• you need real-time mocap for live performance or games
• you require marker-based precision comparable to commercial optical mocap systems
• you want a polished gui application rather than python scripts and research code
• your input is depth sensors or imu data rather than rgb video

Facets

library · maturity active

computer-vision machine-learning image-processing video-processing computer-vision machine-learning graphics python cross-platform motion-capture markerless-mocap smpl human-pose-estimation novel-view-synthesis multi-view 3d-human-reconstruction linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="zju3dv/EasyMocap")