import math import numpy as np import random import torch import torch.utils.data as data import utils.utils_image as util from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels from basicsr.utils import DiffJPEG, USMSharp from numpy.typing import NDArray from PIL import Image from utils.utils_video import img2tensor from torch import Tensor from data.degradations import apply_real_esrgan_degradations class DatasetSR(data.Dataset): ''' # ----------------------------------------- # Get L/H for SISR. # If only "paths_H" is provided, sythesize bicubicly downsampled L on-the-fly. # ----------------------------------------- # e.g., SRResNet # ----------------------------------------- ''' def __init__(self, opt): super(DatasetSR, self).__init__() self.opt = opt self.n_channels = opt['n_channels'] if opt['n_channels'] else 3 self.sf = opt['scale'] if opt['scale'] else 4 self.patch_size = self.opt['H_size'] if self.opt['H_size'] else 96 self.L_size = self.patch_size // self.sf # ------------------------------------ # get paths of L/H # ------------------------------------ self.paths_H = util.get_image_paths(opt['dataroot_H']) self.paths_L = util.get_image_paths(opt['dataroot_L']) assert self.paths_H, 'Error: H path is empty.' if self.paths_L and self.paths_H: assert len(self.paths_L) == len(self.paths_H), 'L/H mismatch - {}, {}.'.format(len(self.paths_L), len(self.paths_H)) self.jpeg_simulator = DiffJPEG() self.usm_sharpener = USMSharp() blur_kernel_list1 = ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso'] blur_kernel_list2 = ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso'] blur_kernel_prob1 = [0.45, 0.25, 0.12, 0.03, 0.12, 0.03] blur_kernel_prob2 = [0.45, 0.25, 0.12, 0.03, 0.12, 0.03] kernel_size = 21 blur_sigma1 = [0.05, 0.2] blur_sigma2 = [0.05, 0.1] betag_range1 = [0.7, 1.3] betag_range2 = [0.7, 1.3] betap_range1 = [0.7, 1.3] betap_range2 = [0.7, 1.3] def _decide_kernels(self) -> NDArray: blur_kernel1 = random_mixed_kernels( self.blur_kernel_list1, self.blur_kernel_prob1, self.kernel_size, self.blur_sigma1, self.blur_sigma1, [-math.pi, math.pi], self.betag_range1, self.betap_range1, noise_range=None ) blur_kernel2 = random_mixed_kernels( self.blur_kernel_list2, self.blur_kernel_prob2, self.kernel_size, self.blur_sigma2, self.blur_sigma2, [-math.pi, math.pi], self.betag_range2, self.betap_range2, noise_range=None ) if self.kernel_size < 13: omega_c = np.random.uniform(np.pi / 3, np.pi) else: omega_c = np.random.uniform(np.pi / 5, np.pi) sinc_kernel = circular_lowpass_kernel(omega_c, self.kernel_size, pad_to=21) return (blur_kernel1, blur_kernel2, sinc_kernel) def __getitem__(self, index): L_path = None # ------------------------------------ # get H image # ------------------------------------ H_path = self.paths_H[index] img_H = util.imread_uint(H_path, self.n_channels) img_H = util.uint2single(img_H) # ------------------------------------ # modcrop # ------------------------------------ img_H = util.modcrop(img_H, self.sf) # ------------------------------------ # get L image # ------------------------------------ if self.paths_L: # -------------------------------- # directly load L image # -------------------------------- L_path = self.paths_L[index] img_L = util.imread_uint(L_path, self.n_channels) img_L = util.uint2single(img_L) else: # -------------------------------- # sythesize L image via matlab's bicubic # -------------------------------- H, W = img_H.shape[:2] img_L = util.imresize_np(img_H, 1 / self.sf, True) src_tensor = img2tensor(img_L.copy(), bgr2rgb=False, float32=True).unsqueeze(0) blur_kernel1, blur_kernel2, sinc_kernel = self._decide_kernels() (img_L_2, sharp_img_L, degraded_img_L) = apply_real_esrgan_degradations( src_tensor, blur_kernel1=Tensor(blur_kernel1).unsqueeze(0), blur_kernel2=Tensor(blur_kernel2).unsqueeze(0), second_blur_prob=0.2, sinc_kernel=Tensor(sinc_kernel).unsqueeze(0), resize_prob1=[0.2, 0.7, 0.1], resize_prob2=[0.3, 0.4, 0.3], resize_range1=[0.9, 1.1], resize_range2=[0.9, 1.1], gray_noise_prob1=0.2, gray_noise_prob2=0.2, gaussian_noise_prob1=0.2, gaussian_noise_prob2=0.2, noise_range=[0.01, 0.2], poisson_scale_range=[0.05, 0.45], jpeg_compression_range1=[85, 100], jpeg_compression_range2=[85, 100], jpeg_simulator=self.jpeg_simulator, random_crop_gt_size=256, sr_upsample_scale=1, usm_sharpener=self.usm_sharpener ) # Image.fromarray((degraded_img_L[0] * 255).permute( # 1, 2, 0).cpu().numpy().astype(np.uint8)).save( # "/home/cll/Desktop/degraded_L.png") # Image.fromarray((img_L * 255).astype(np.uint8)).save( # "/home/cll/Desktop/img_L.png") # Image.fromarray((img_L_2[0] * 255).permute( # 1, 2, 0).cpu().numpy().astype(np.uint8)).save( # "/home/cll/Desktop/img_L_2.png") # exit() # ------------------------------------ # if train, get L/H patch pair # ------------------------------------ if self.opt['phase'] == 'train': H, W, C = img_L.shape # -------------------------------- # randomly crop the L patch # -------------------------------- rnd_h = random.randint(0, max(0, H - self.L_size)) rnd_w = random.randint(0, max(0, W - self.L_size)) img_L = img_L[rnd_h:rnd_h + self.L_size, rnd_w:rnd_w + self.L_size, :] # -------------------------------- # crop corresponding H patch # -------------------------------- rnd_h_H, rnd_w_H = int(rnd_h * self.sf), int(rnd_w * self.sf) img_H = img_H[rnd_h_H:rnd_h_H + self.patch_size, rnd_w_H:rnd_w_H + self.patch_size, :] # -------------------------------- # augmentation - flip and/or rotate + RealESRGAN modified degradations # -------------------------------- mode = random.randint(0, 7) img_L, img_H = util.augment_img(img_L, mode=mode), util.augment_img(img_H, mode=mode) # ------------------------------------ # L/H pairs, HWC to CHW, numpy to tensor # ------------------------------------ img_H, img_L = util.single2tensor3(img_H), util.single2tensor3(img_L) if L_path is None: L_path = H_path return {'L': img_L, 'H': img_H, 'L_path': L_path, 'H_path': H_path} def __len__(self): return len(self.paths_H)