yohanshin/WHAM

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1130 stars · 141 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

WHAM is the official PyTorch implementation of the CVPR 2024 paper 'Reconstructing World-grounded Humans with Accurate 3D Motion'. It estimates 3D human pose and motion from monocular video, recovering world-grounded body motion using SMPL body models with optional SLAM and camera calibration.

Use cases

• reconstruct 3d human motion from a single video
• estimate world-grounded 3d human pose from monocular video
• run human motion capture on smartphone videos
• get smpl body model parameters from video
• estimate camera-coordinate human motion without slam
• refine 3d human pose estimates with temporal smplify

When to choose

• you need state-of-the-art world-grounded 3D human pose and motion from monocular video
• you want SMPL body model outputs for downstream animation or analysis
• you have videos with known or estimable camera intrinsics and want global motion via SLAM

When to avoid

• you need real-time pose estimation on edge devices
• you cannot register to download the SMPL/SMPLify body models
• you need multi-person reconstruction in crowded scenes out of the box

Facets

library · maturity active

computer-vision machine-learning video-processing computer-vision machine-learning artificial-intelligence python 3d-human-pose-estimation smpl human-motion-capture pytorch research-code slam video linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="yohanshin/WHAM")