Continual All-in-One Adverse Weather Removal with Knowledge Replay on a Unified Network Structure [TMM 2024]

[Continual All-in-One Adverse Weather Removal with Knowledge Replay on a Unified Network Structure]
[De Cheng], [Yanling Ji], [Dong Gong], [Nannan Wang], [Junwei Han], [Dingwen Zhang]

Requirements

Python 3.6+
pip install -r requirements.txt

Experimental Setup

Our code requires three datasets: OTS, Rain100H, Snow100K

Dataset

We recommend putting all datasets under the same folder (say $datasets) to ease management and following the instructions below to organize datasets to avoid modifying the source code. The file structure should look like:

$datasets/
|–– RESIDE/
    |–– OTS_beta/
        |–– hazy/
        |–– clear/
    |–– SOTS/
        |–– outdoor/
            |–– hazy/
            |–– clear/
|–– Rain100H/
    |–– train
        |–– rain
        |–– norain
    |–– test
        |–– rain
        |–– norain
|–– Snow100K
    |–– train
        |–– synthetic
        |–– gt
    |–– test
        |–– Snow100K-M
            |–– synthetic
            |–– gt

First, run python patch.py to patch Rain100H.

Usage

If you want to reproduce the results mentioned in our paper, run

python main.py --task_order haze rain snow --memory_size 500 --exp_name haze_rain_snow --eval_step 20000 --device cuda:0

We provide our training checkpoints and you can continue training using the --resume hyperparameter.

Hyperparameters

The meaning of hyperparameters in the command line is as follows:

params	name
--task_order	task order for dehazing, deraining, desnowing
--memory_size	memory size
--exp_name	experiment name
--eval_step	each step for eval
--resume	training from previous tasks

If you encounter any issues or have any questions, please let us know.