# elliottwu/unsup3d

(CVPR'20 Oral) Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild

Repository: https://github.com/elliottwu/unsup3d
Canonical: https://ross.abutalabs.com/products/unsup3d
Language: Python
License: MIT
License Family: permissive
Topics: 3d-objects, 3d, 3d-reconstuction, unsupervised, unsupervised-learning, cvpr, cvpr-2020, cvpr2020, pytorch
Last push: 2021-08-31T03:12:23+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2367, "days_push": 1828, "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 1189, forks 191 (observed 2026-08-28T04:03:55.825142+00:00)

## What it is
Official PyTorch implementation of the CVPR 2020 (Oral, Best Paper Award) research paper 'Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild' from Oxford's Visual Geometry Group. It learns 3D shape, pose, albedo, and lighting of deformable object categories (e.g., faces, cats) from raw single-view images without any 3D ground truth or keypoint supervision.

## Use cases
- reconstruct 3D objects from single images without 3D ground truth
- learn 3D face shape from unannotated photos
- run unsupervised 3D reconstruction research experiments
- reproduce the Unsup3D CVPR 2020 paper results
- estimate depth, albedo, and lighting from a single image
- train a 3D-aware model on cat or human face datasets

## When to choose
- you need a research baseline for unsupervised single-view 3D reconstruction
- you want to reproduce or extend the CVPR 2020 Unsup3D paper
- you have single-view image collections of symmetric deformable objects like faces
- you need a reference implementation with a pretrained demo

## When to avoid
- you need production-ready, actively maintained 3D reconstruction software
- you lack a GPU, since training and testing require GPU-only neural rendering
- you need modern PyTorch versions, as the code targets PyTorch 1.2.0 and CUDA 9.2
- your objects are not roughly symmetric or deformable

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: 3d-reconstruction, unsupervised-learning, pytorch, cvpr-2020, single-view-3d, deformable-objects, research-code, neural-rendering, linux, gpu

## Member repositories
- elliottwu/unsup3d (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:55.825142+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 2026-08-30T06:23:08.836570+00:00, confidence not recorded.
  - readme: https://github.com/elliottwu/unsup3d (fetched 2026-08-28T04:03:55.825142+00:00, sha 0943fb64e230)
- Data as of 2026-08-30T08:39:29.467469+00:00.