# zju3dv/GVHMR

Code for "GVHMR: World-Grounded Human Motion Recovery via Gravity-View Coordinates", Siggraph Asia 2024, TPAMI 2026

Repository: https://github.com/zju3dv/GVHMR
Canonical: https://ross.abutalabs.com/products/gvhmr
Homepage: https://zju3dv.github.io/gvhmr
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Last push: 2026-05-21T12:27:46+00:00

## Health v2 (maintenance only)
Score: 60/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 83, release rhythm 35, longevity 52
- inputs: {"age_days": 728, "days_push": 104, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1882, forks 236 (observed 2026-08-28T04:05:48.676864+00:00)

## What it is
GVHMR is a research codebase implementing the SIGGRAPH Asia 2024 paper 'World-Grounded Human Motion Recovery via Gravity-View Coordinates'. It recovers world-grounded global 3D human motion (SMPL poses) from monocular video using a gravity-view coordinate representation, with training and inference pipelines.

## Use cases
- recover 3d human motion from monocular video
- estimate world-grounded smpl poses from video
- run human pose estimation on a video with a static or moving camera
- reproduce 3dpw rich and emdb benchmark results
- train a human motion recovery model on amass bedlam datasets
- demo human motion capture from a single video

## When to choose
- you need global (world-grounded) human motion rather than per-frame camera-relative pose
- you want to reproduce or build on a published state-of-the-art motion recovery method
- you have a GPU and want research-grade training and inference code

## When to avoid
- you need a production-ready commercial product (license is non-commercial)
- you need real-time motion capture on CPU-only hardware
- you only need simple 2D pose estimation

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, computer-vision, deep-learning, simulation
- domain: computer-vision, machine-learning, artificial-intelligence
- platform: python
- tags: human-motion-recovery, pose-estimation, smpl, monocular-video, pytorch, 3d-human-pose, world-grounded-motion, research-code, gpu, linux

## Member repositories
- zju3dv/GVHMR (main) score 60

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:48.676864+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:13:57.485414+00:00, confidence not recorded.
  - readme: https://github.com/zju3dv/GVHMR (fetched 2026-08-28T04:05:48.676864+00:00, sha 0c4b6b619bcf)
  - homepage: https://zju3dv.github.io/gvhmr (fetched 2026-08-29T10:53:02.251478+00:00, sha 763e52f4169b)
  - registry_pypi: https://pypi.org/pypi/gvhmr/json (fetched 2026-08-29T10:53:02.260496+00:00, sha 2b58302d01f0)
- Data as of 2026-08-30T08:39:29.467469+00:00.