Need help with noise2noise-pytorch?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

164 Stars 39 Forks MIT License 154 Commits 6 Opened issues


PyTorch Implementation of Noise2Noise (Lehtinen et al., 2018)

Services available


Need anything else?

Contributors list

# 232,085
Deep le...
141 commits

Noise2Noise: Learning Image Restoration without Clean Data

This is an unofficial PyTorch implementation of Noise2Noise (Lehtinen et al. 2018).


To install the latest version of all packages, run

pip3 install --user -r requirements.txt

This code was tested on Python 3.6.5 on macOS High Sierra (10.13.4) and Ubuntu 16.04. It will fail with Python 2.7.x due to usage of 3.6-specific functions. Note that training and testing will also fail on Windows out of the box due to differences in the path resolvers (



The authors use ImageNet, but any dataset will do. COCO 2017 has a small validation set (1 GB) which can be nicely split into train/valid for easier training. For instance, to obtain a 4200/800 train/valid split you can do:

mkdir data && cd data
mkdir train valid test
unzip && cd val2017
mv `ls | head -4200` ../train
mv `ls | head -800` ../valid

You can also download the full datasets (7 GB) that more or less match the paper, if you have the bandwidth:

mkdir data && cd data
mkdir train valid test
unzip -j -d train
unzip -j -d valid

Add your favorite images to the

folder. Only a handful will do to visually inspect the denoiser performance.



python3 --h
for list of optional arguments, or
for an example.

By default, the model train with noisy targets. To train with clean targets, use

. To train and validate on smaller datasets, use the
options. To plot stats as the model trains, use
; these are saved alongside checkpoints. By default CUDA is not enabled: use the
option if you have a GPU that supports it.

Gaussian noise

The noise parameter is the maximum standard deviation σ.

python3 \
  --train-dir ../data/train --train-size 1000 \
  --valid-dir ../data/valid --valid-size 200 \
  --ckpt-save-path ../ckpts \
  --nb-epochs 10 \
  --batch-size 4 \
  --loss l2 \
  --noise-type gaussian \
  --noise-param 50 \
  --crop-size 64 \
  --plot-stats \

Poisson noise

The noise parameter is the Poisson parameter λ.

  --loss l2 \
  --noise-type poisson \
  --noise-param 50 \

Text overlay

The noise parameter is the approximate probability p that a pixel is covered by text.

python3 \
  --loss l1 \
  --noise-type text \
  --noise-param 0.5 \

Monte Carlo rendering noise

See other read-me file.


Model checkpoints are automatically saved after every epoch. To test the denoiser, provide
with a PyTorch model (
file) via the argument
and a test image directory via
. The
option specifies the number of noisy/denoised/clean montages to display on screen. To disable this, simply remove
python3 \
  --data ../data \
  --load-ckpt ../ckpts/gaussian/ \
  --noise-type gaussian \
  --noise-param 50 \
  --crop-size 256 \
  --show-output 3 \


python3 --h
for list of optional arguments, or
for an example.


Gaussian model was trained for 100 epochs with a train/valid split of 2000/400. Text model was trained with 1000/200 due to the overhead introduced when corrupting with text. Both models were trained on an old NVIDIA GTX 780.

Gaussian noise (σ = 25)
Noisy input (20.34 dB) Denoised (32.68 dB) Clean targets (32.49 dB) Ground truth
Text overlay (p = 0.25)
Noisy input (15.07 dB) Denoised (28.10 dB) Clean targets (27.79 dB) Ground truth

Known issues

  • U-Net has shared weights (i.e. same
    for the decoder part). This is trivial to fix but I simply don't have the time to do so. The model seems to perform well even with this error.
  • Activation functions should be LeakyReLUs everywhere except last layer where it should be ReLU. The current implementation (and pretrained models) assumes the opposite due to me originally misreading the paper. It was reported that using the correct version yields unstable training without batch norm; I haven't tested it yet.
  • It is unclear how to deal with Poisson noise since it is data-dependent and thus nonadditive. See official TensorFlow implementation to adapt properly.



I would like to acknowledge Yusuke Uchida for his Keras implementation of Noise2Noise. Although Keras and PyTorch are very different frameworks, parts of his code did help me in completing this implementation.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.