4ever-rain / deraindrop_unofficial Goto Github PK

Reference: https://github.com/rui1996/DeRaindrop

Fork from https://github.com/shleecs/DeRaindrop_unofficial . Try to change the GPUs' problems.

1. Linux
2. Python 3.7
3. NVIDIA GPU + CUDA CuDNN (CUDA 10.0)
4. Pytorch 1.4.1
5. TensorboardX

python train.py --gpu 0,1 --train_dataset ../dataset/Derain/train/train/ --eval_dataset ../dataset/Derain/test_a/test_a/ ../dataset/Derain/test_b/test_b/ --batch_size 4

The demo pictures are put under the directory ./demo/input/ and ./demo/output/ is a sample of the ouput of the model. If you want to generate your own, use the following code:

CUDA_VISIBLE_DEVICES=gpu_id python predict.py --mode demo --input_dir ./demo/input/ --output_dir ./demo/output/

Please change the line 60 of predict.py to your own G.pth dir

Training source / loss implementation / dataloader ...

The whole dataset can be find in author pages(https://github.com/rui1996/DeRaindrop)