|
- """
- ## Multi-Stage Progressive Image Restoration
- ## Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao
- ## https://arxiv.org/abs/2102.02808
- """
-
- import numpy as np
- import os
- import argparse
- from tqdm import tqdm
-
- import torch
- import torch.nn as nn
- import torch.nn.functional as F
- import utils
-
- from MPRNet import MPRNet
- from skimage import img_as_ubyte
- import h5py
- import scipy.io as sio
- from pdb import set_trace as stx
-
- parser = argparse.ArgumentParser(description='Image Denoising using MPRNet')
-
- parser.add_argument('--input_dir', default='./Datasets/SIDD/test/', type=str, help='Directory of validation images')
- parser.add_argument('--result_dir', default='./results/SIDD/', type=str, help='Directory for results')
- parser.add_argument('--weights', default='./pretrained_models/model_denoising.pth', type=str, help='Path to weights')
- parser.add_argument('--gpus', default='1', type=str, help='CUDA_VISIBLE_DEVICES')
- parser.add_argument('--save_images', action='store_true', help='Save denoised images in result directory')
-
- args = parser.parse_args()
-
- os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
- os.environ["CUDA_VISIBLE_DEVICES"] = args.gpus
-
- result_dir = os.path.join(args.result_dir, 'mat')
- utils.mkdir(result_dir)
-
- if args.save_images:
- result_dir_img = os.path.join(args.result_dir, 'png')
- utils.mkdir(result_dir_img)
-
- model_restoration = MPRNet()
-
- utils.load_checkpoint(model_restoration,args.weights)
- print("===>Testing using weights: ",args.weights)
- model_restoration.cuda()
- model_restoration = nn.DataParallel(model_restoration)
- model_restoration.eval()
-
- # Process data
- filepath = os.path.join(args.input_dir, 'ValidationNoisyBlocksSrgb.mat')
- img = sio.loadmat(filepath)
- Inoisy = np.float32(np.array(img['ValidationNoisyBlocksSrgb']))
- Inoisy /=255.
- restored = np.zeros_like(Inoisy)
- with torch.no_grad():
- for i in tqdm(range(40)):
- for k in range(32):
- noisy_patch = torch.from_numpy(Inoisy[i,k,:,:,:]).unsqueeze(0).permute(0,3,1,2).cuda()
- restored_patch = model_restoration(noisy_patch)
- restored_patch = torch.clamp(restored_patch[0],0,1).cpu().detach().permute(0, 2, 3, 1).squeeze(0)
- restored[i,k,:,:,:] = restored_patch
-
- if args.save_images:
- save_file = os.path.join(result_dir_img, '%04d_%02d.png'%(i+1,k+1))
- utils.save_img(save_file, img_as_ubyte(restored_patch))
-
- # save denoised data
- sio.savemat(os.path.join(result_dir, 'Idenoised.mat'), {"Idenoised": restored,})
|