About this repo:

Unofficial pytorch implementation for TransWeather -- CVPR 2022
Original paper : TransWeather: Transformer-based Restoration of Images Degraded by Adverse Weather Conditions
Note : Most of the code in this repository comes from the original author's repository. Here, it is more about fine-tuning the source code, adding a lot of comments, fixing some existing bugs(or improve the performance), and adding some other details. (At the time of completing this repository, some details of the original author's code were not completed, so if the original author has an update, I will continue to follow up.)
Plus, the original paper uses a larger dataset, which is inconvenient for people who just learn the model architecture and run the code successfully. So I sampled 600 images from the original dataset, 480 for training and 120 for testing, which you can download here:Allweather_subset.zip, it's only 170mb.

Introduction

Name		Name	Last commit message	Last commit date
Latest commit History 103 Commits
imgs		imgs
models		models
scripts		scripts
utils		utils
LICENSE		LICENSE
README.md		README.md
environment.yml		environment.yml
evaluate_and_visual.py		evaluate_and_visual.py
requirements.txt		requirements.txt
test.py		test.py
test_raindrop.py		test_raindrop.py
test_raindropa.py		test_raindropa.py
test_snow100k.py		test_snow100k.py
test_test1.py		test_test1.py
train_ddp_swtrans.py		train_ddp_swtrans.py
train_gpu_swtrans.py		train_gpu_swtrans.py
train_gpu_trans.py		train_gpu_trans.py
train_individual_trans.py		train_individual_trans.py
train_original_trans.py		train_original_trans.py