zaghlol94 / noise2sim

Noise2Sim -- Similarity-based Self-Learning for Image Denoising

Under review.

Introduction

This project is the Pytorch implementation of the paper, and was developed based on our clustering project.

Installation

Assuming Anaconda with python 3.6, the required packages for this project can be installed as:

conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch
conda install faiss-gpu cudatoolkit=10.0 -c pytorch
conda install matplotlib
conda install -c conda-forge python-lmdb tqdm imageio addict tensorboard opencv

Then, clone this repo

git clone https://github.com/niuchuangnn/noise2sim.git
cd noise2sim

Run

To train the model, simply run the following commands.

Download BSD68 test dataset

python ./tools/download_bsd68_noise2void.py

Prepare dataset:

python ./tools/prepare_bsd400_lmdb.py

Run on 1 GPU:

python ./tools/train_dist.py --config-file ./configs/bsd400_unet2_ps3_ns8_gpu1.py

Run on 8 GPUs:

python ./tools/train_dist.py --config-file ./configs/bsd400_unet2_ps3_ns8_gpu8.py

The results in paper were obtained using 8 GPUs, you can obtain similar results with 1 GPU.

TODO

More experiments.

Citation

@inproceedings{noise2sim2021,
  title={Noise2Sim – Similarity-based Self-Learning for Image Denoising},
  author={Niu, Chuang and Wang, Ge},
  booktitle={arXiv:2011.03384},
  year={2020}
}