# yu4u/noise2noise

An unofficial and partial Keras implementation of "Noise2Noise: Learning Image Restoration without Clean Data"

Repository: https://github.com/yu4u/noise2noise
Canonical: https://ross.abutalabs.com/products/yu4u-noise2noise
Language: Python
License: MIT
License Family: permissive
Topics: noise2noise, deep-learning, convolutional-neural-networks, denoising, keras
Last push: 2021-08-12T12:14:29+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 2967, "days_push": 1847, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1088, forks 233 (observed 2026-08-28T04:03:32.458079+00:00)

## What it is
An unofficial, partial Keras implementation of the Noise2Noise paper for learning image restoration without clean training data. It trains denoising models (SRResNet or UNet) using pairs of noisy images with Gaussian, text, and impulse noise models.

## Use cases
- denoise images without clean reference data
- train a noise2noise denoising model in Keras
- experiment with L0 loss for impulse noise removal
- compare noise2noise training vs standard clean-target training
- remove text overlays from images with a CNN

## When to choose
- you want a simple, readable Keras implementation of noise2noise for learning or research
- you need to train denoising models on noisy-noise image pairs
- you work in the Keras/TensorFlow ecosystem

## When to avoid
- you need the exact RED30 model or full reproduction of the paper
- you need a maintained, production-grade denoising library
- you use PyTorch or need modern TensorFlow 2.x support

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing
- domain: machine-learning, computer-vision, image-processing
- platform: python
- tags: noise2noise, denoising, keras, tensorflow, image-restoration, unofficial-implementation

## Member repositories
- yu4u/noise2noise (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:32.458079+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 2026-08-30T06:49:24.340452+00:00, confidence not recorded.
  - readme: https://github.com/yu4u/noise2noise (fetched 2026-08-28T04:03:32.458079+00:00, sha f9353b395743)
- Data as of 2026-08-30T08:39:29.467469+00:00.