yanto-manta / dctgan-master

GAN-Based Image Deblurring Using DCT Discriminator

GAN-Based Image Deblurring Using DCT Discriminator

Installation

Clone this project to your machine.

git clone https://github.com/Hiroki-Tomosada/DCTGAN-master.git
cd DCTGAN-master/deblur/

Dataset

The datasets for training can be downloaded via the links below:

• GOPRO
• DVD
• NFS
• HIDE

Train

python main-Pytorch.py --data_folder <MODEL_PATH> --dataset_train <DATASET_PATH>

Test

GOPRO

python main-Pytorch.py --data_folder <MODEL_PATH> --test_dataset "<DATASET_PATH>/GOPRO_Large/test/*/blur/*.*" --result <RESULT_PATH> --test

DVD

python main-Pytorch.py --data_folder <MODEL_PATH> --test_dataset "<DATASET_PATH>/DVD_3840FPS_AVG_3-21/test/*/*.png" --result <RESULT_PATH> --test

NFS

python main-Pytorch.py --data_folder <MODEL_PATH> --test_dataset "<DATASET_PATH>/NFS_3840FPS_AVG_3-21/test/*/*.png" --result <RESULT_PATH> --test

HIDE

python main-Pytorch.py --data_folder <MODEL_PATH> --test_dataset "<DATASET_PATH>/HIDE_dataset/test/*/*.png" --result <RESULT_PATH> --test

real_dataset

python main-Pytorch.py --data_folder <MODEL_PATH> --test_dataset "<DATASET_PATH>/real_dataset/*.jpg" --result <RESULT_PATH> --test

Kernel Prediction

Train

python main-Pytorch.py --gpu_id 0 --data_folder <MODEL_PATH> --model Kernel-Prediction --epoch 100 --pretrain 100 --dataset MSCOCO --dataset_train <DATASET_PATH>/MSCOCO/train_apply2 --image_size 100 --kernel_mode Estimate-Kernel-size --kernel_min 2 --kernel_max 20

Predict

GOPRO

python main-Pytorch.py --gpu_id 0 --data_folder <MODEL_PATH> --model Kernel-Prediction --test --test_dataset "<DATASET_PATH>/GOPRO_Large/train/*/*blur/*.*" --result <RESULT_PATH>/GOPRO_kernel  --image_size 100 --kernel_mode Estimate-Kernel-size --patch

DVD

python main-Pytorch.py --gpu_id 0 --data_folder <MODEL_PATH> --model Kernel-Prediction --test --test_dataset "<DATASET_PATH>/DVD_3840FPS_AVG_3-21/train/blur/*/*.png" --result <RESULT_PATH>/DVD_kernel  --image_size 100 --kernel_mode Estimate-Kernel-size --patch

NFS

python main-Pytorch.py --gpu_id 0 --data_folder <MODEL_PATH> --model Kernel-Prediction --test --test_dataset "<DATASET_PATH>/NFS_3840FPS_AVG_3-21/train/blur/*/*.png" --result <RESULT_PATH>/NFS_kernel  --image_size 100 --kernel_mode Estimate-Kernel-size --patch

HIDE

python main-Pytorch.py --gpu_id 0 --data_folder <MODEL_PATH> --model Kernel-Prediction --test --test_dataset "<DATASET_PATH>/HIDE_dataset/train/*.png" --result <RESULT_PATH>/HIDE_kernel  --image_size 100 --kernel_mode Estimate-Kernel-size --patch

Make Dataset

python make_dataset.py --mode Blur-Classification-All --save_path "<DATASET_PATH>/Kernel_dataset/Customized" --theory --average 14 --sigma 1 --min 0 --max 30