# MoyGcc/vid2avatar

Vid2Avatar: 3D Avatar Reconstruction from Videos in the Wild via Self-supervised Scene Decomposition (CVPR2023)

Repository: https://github.com/MoyGcc/vid2avatar
Canonical: https://ross.abutalabs.com/products/vid2avatar
Homepage: https://moygcc.github.io/vid2avatar/
Language: Python
License: MIT
License Family: permissive
Topics: 3d-human-reconstruction
Last push: 2025-12-17T15:47:53+00:00

## Health v2 (maintenance only)
Score: 56/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 57, release rhythm 35, longevity 92
- inputs: {"age_days": 1299, "days_push": 259, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1360, forks 110 (observed 2026-08-28T04:04:29.932740+00:00)

## What it is
Vid2Avatar is the official PyTorch implementation of a CVPR 2023 method that reconstructs detailed 3D human avatars from monocular in-the-wild videos. It jointly models the human and background as separate neural fields, self-supervisedly decomposing the scene to produce canonical human geometry and texture without ground-truth supervision or external segmentation.

## Use cases
- reconstruct a 3D human avatar from a single monocular video
- separate a moving person from an arbitrary background in 3D
- generate canonical and deformed human mesh sequences from video
- render novel views of a reconstructed clothed human
- research baseline for neural human avatar reconstruction
- visualize SMPL-based human models in 3D

## When to choose
- you need a self-supervised 3D human reconstruction pipeline from short monocular clips
- you want a research-grade CVPR 2023 baseline for neural avatar creation
- you have GPU resources and can tolerate 24-48 hour per-sequence training
- you want canonical-space human geometry and texture without labeled data

## When to avoid
- you need real-time or fast avatar reconstruction
- you want a polished end-user application rather than research code
- you cannot obtain the SMPL body model or set up a CUDA/conda environment
- your input is multi-view studio footage rather than monocular in-the-wild video

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, computer-vision, image-processing, graphics, simulation
- domain: computer-vision, deep-learning, graphics, machine-learning
- platform: python
- tags: 3d-reconstruction, neural-fields, human-avatar, smpl, cvpr-2023, research-code, monocular-video, scene-decomposition, linux, gpu

## Member repositories
- MoyGcc/vid2avatar (main) score 56

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:29.932740+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 2026-08-30T04:41:37.219929+00:00, confidence not recorded.
  - readme: https://github.com/MoyGcc/vid2avatar (fetched 2026-08-28T04:04:29.932740+00:00, sha a3981b374f4a)
  - homepage: https://moygcc.github.io/vid2avatar/ (fetched 2026-08-29T11:59:24.520804+00:00, sha 5e8455733211)
- Data as of 2026-08-30T08:39:29.467469+00:00.