# tamarott/SinGAN

Official pytorch implementation of the paper: "SinGAN: Learning a Generative Model from a Single Natural Image"

Repository: https://github.com/tamarott/SinGAN
Canonical: https://ross.abutalabs.com/products/singan
Homepage: https://tamarott.github.io/SinGAN.htm
Language: Python
License: NOASSERTION
License Family: other
Topics: singan, gan, official, single-image, harmonization, animation, single-image-animation, single-image-generation, image-edit, arbitrery-sizes, single-image-super-resolution, super-resolution
Last push: 2023-05-26T07:19:48+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": 2573, "days_push": 1195, "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 3344, forks 621 (observed 2026-08-28T04:07:56.983989+00:00)

## What it is
Official PyTorch implementation of SinGAN, an ICCV 2019 best-paper generative model trained on a single natural image. It learns patch statistics across scales to generate random samples and supports image manipulation tasks like harmonization, editing, super-resolution, paint-to-image, and animation.

## Use cases
- train a generative model from a single image
- generate random image samples of arbitrary size
- super-resolution of a single image
- image harmonization of pasted objects
- turn a paint or clipart into a realistic photo
- animate a single image
- edit and rearrange objects in a photo

## When to choose
- you want to generate diverse samples from just one training image without a large dataset
- you need single-image super-resolution or harmonization with a research-grade reference implementation
- you are reproducing the ICCV 2019 SinGAN paper results

## When to avoid
- you need modern PyTorch support - the code only works with torch 1.4 or earlier
- you need large-scale or fast image generation, since a model must be trained per image
- you want a maintained production tool rather than research code

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing, stable-diffusion
- domain: deep-learning, computer-vision, image-processing, artificial-intelligence
- platform: python, windows
- tags: gan, single-image-generation, super-resolution, image-harmonization, image-animation, pytorch, research-code, iccv-2019, linux, macos, gpu

## Member repositories
- tamarott/SinGAN (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:56.983989+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 2026-08-29T18:41:02.433061+00:00, confidence not recorded.
  - readme: https://github.com/tamarott/SinGAN (fetched 2026-08-28T04:07:56.983989+00:00, sha 05fd675e6b32)
  - homepage: https://tamarott.github.io/SinGAN.htm (fetched 2026-08-29T09:34:22.657896+00:00, sha bb7290956e4d)
- Data as of 2026-08-30T08:39:29.467469+00:00.