# mit-han-lab/data-efficient-gans

[NeurIPS 2020] Differentiable Augmentation for Data-Efficient GAN Training

Repository: https://github.com/mit-han-lab/data-efficient-gans
Canonical: https://ross.abutalabs.com/products/data-efficient-gans
Homepage: https://arxiv.org/abs/2006.10738
Language: Python
License: BSD-2-Clause
License Family: permissive
Topics: gans, pytorch, tensorflow, data-efficient, generative-adversarial-network, image-generation, neurips-2020
Last push: 2024-09-24T17:46:07+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2270, "days_push": 708, "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 1308, forks 173 (observed 2026-08-28T04:04:19.149540+00:00)

## What it is
Official implementation of Differentiable Augmentation (DiffAugment), a NeurIPS 2020 method that improves GAN training data efficiency by applying differentiable augmentations to both real and generated samples. It includes StyleGAN2 and BigGAN variants in PyTorch and TensorFlow for low-shot and limited-data image generation.

## Use cases
- train a GAN with only 100 images
- generate images from a small dataset without pre-training
- improve GAN training stability with limited data
- reproduce StyleGAN2 results on CIFAR-10 with less data
- apply differentiable augmentation to GAN training
- train BigGAN on ImageNet with TPU

## When to choose
- you have very few training images (100-1000) and want high-fidelity GAN generation
- you want a research-validated augmentation method compatible with StyleGAN2 or BigGAN
- you need both PyTorch and TensorFlow implementations

## When to avoid
- you need diffusion models or modern generative approaches instead of GANs
- you have abundant training data where standard GANs already work well
- you need actively maintained production software rather than research code

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, image-processing, deep-learning
- domain: deep-learning, computer-vision, image-processing, artificial-intelligence
- platform: python, cross-platform
- tags: gan, diffaugment, stylegan2, biggan, data-augmentation, image-generation, pytorch, tensorflow, research-code, neurips-2020, gpu, linux

## Member repositories
- mit-han-lab/data-efficient-gans (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:19.149540+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 2026-08-30T04:50:05.330245+00:00, confidence not recorded.
  - readme: https://github.com/mit-han-lab/data-efficient-gans (fetched 2026-08-28T04:04:19.149540+00:00, sha 54aff33b20db)
  - homepage: https://arxiv.org/abs/2006.10738 (fetched 2026-08-29T12:08:31.211645+00:00, sha b42aef2d1729)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T12:08:31.220954+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T12:08:31.224186+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T12:08:31.225885+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T12:08:31.222651+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.