# autonomousvision/giraffe

This repository contains the code for the CVPR 2021 paper "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields"

Repository: https://github.com/autonomousvision/giraffe
Canonical: https://ross.abutalabs.com/products/autonomousvision-giraffe
Homepage: https://m-niemeyer.github.io/project-pages/giraffe/index.html
Language: Python
License: MIT
License Family: permissive
Topics: cvpr2021, generative-model, generative-modelling, generative-adversarial-network, nerf, implicit-surfaces, neural-scene-representations
Last push: 2022-02-09T08:06:14+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": 1986, "days_push": 1666, "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 1238, forks 156 (observed 2026-08-28T04:04:05.448504+00:00)

## What it is
Official PyTorch implementation of GIRAFFE, a CVPR 2021 (oral, best paper award) generative model that represents scenes as compositional generative neural feature fields. It combines a compositional 3D scene representation with neural rendering to enable controllable, photorealistic image synthesis from unposed image collections.

## Use cases
- generate controllable images of scenes with movable objects
- train a GAN with compositional 3D scene representation
- render pre-trained models on cars or CelebA-HQ datasets
- translate and rotate objects in generated scenes
- change camera pose in generated images
- research NeRF-style implicit neural scene representations
- reproduce CVPR 2021 paper results

## When to choose
- you need controllable image synthesis with disentangled objects, shapes, and appearances
- you are researching compositional 3D-aware generative models or NeRF variants
- you want to render or fine-tune provided pre-trained models on standard datasets

## When to avoid
- you need a production-ready image generation service rather than research code
- you lack a GPU or cannot manage the conda environment dependencies
- you need 2D-only image generation without 3D scene control
- the project is research code from 2021 and may need adaptation for modern frameworks

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing, graphics
- domain: computer-vision, deep-learning, machine-learning, graphics
- platform: python
- tags: gan, nerf, generative-model, 3d-scene-representation, controllable-image-synthesis, cvpr2021, research-code, linux, gpu

## Member repositories
- autonomousvision/giraffe (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:05.448504+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 2026-08-30T05:10:02.688232+00:00, confidence not recorded.
  - readme: https://github.com/autonomousvision/giraffe (fetched 2026-08-28T04:04:05.448504+00:00, sha 1e0d2e646176)
  - homepage: https://m-niemeyer.github.io/project-pages/giraffe/index.html (fetched 2026-08-29T12:21:07.135888+00:00, sha e1a16d77821a)
- Data as of 2026-08-30T08:39:29.467469+00:00.