In this repository we provide code of the paper:

MANet: Improving Video Denoising with a Multi-Alignment Network

Yaping Zhao, Haitian Zheng, Zhongrui Wang, Jiebo Luo, Edmund Y. Lam

arxiv link: https://arxiv.org/abs/2202.09704

Input	Output

Flow 1	Flow 2	Flow 3	Flow 4

If you simply want a quick start without training or downloading the model, try running:

conda env create -f environment.yml
conda activate manet
python train_FlowNet_ms_warping_multi.py --test_or_train train

0. For pre-requisites, run:

conda env create -f environment.yml
conda activate manet

1. Pretrained model is currently available at Google Drive and Baidu Netdisk (password: tof6), download the 11.pth and put it in the snapshots/multi_flow...batchsize_1/ folder.

2. For MANet training, run:

python train_FlowNet_ms_warping_multi.py  \
--batch_size_train 1 \
--batch_size_test 1  \
--test_or_train train \
--test_snapshot_file 11.pth \
--cuda_devices 0

This repository only uses several training sample for demo purpose. If you want to train your own model, please prepare your own training set.

3. For MANet testing, run:

python train_FlowNet_ms_warping_multi.py  \
--batch_size_train 1 \
--batch_size_test 1  \
--test_or_train test \
--test_snapshot_file 11.pth \
--cuda_devices 0

This repository only uses several testing sample for demo purpose. If you want to train your own model, please prepare your own training set.

4. If positive, you will get results in the snapshots/[model_name]/visualization_train/ folder when training and results in the snapshots/[model_name]/visualization_test/ folder when testing.

Dataset is stored in the folder dataset/, where subfolder vimeo_septuplet/ contains only several samples for demo purpose. Images in sequences/ and sequences_noise are ground truth clean images and noisy images, respectively.

sep_testlist.txt and sep_trainlist.txt could be modified for your experiment on other datasets.

Cite our paper if you find it interesting!

@article{zhao2022manet,
  title={MANet: Improving Video Denoising with a Multi-Alignment Network},
  author={Zhao, Yaping and Zheng, Haitian and Wang, Zhongrui and Luo, Jiebo and Lam, Edmund Y},
  journal={arXiv preprint arXiv:2202.09704},
  year={2022}
}