Datasets:

AaronCIH
/

PIR_tar

Modalities:

PIR_tar / scripts /generate_lowresolution.py

Rename generate_lowresolution.py to scripts/generate_lowresolution.py

62abb2b verified about 1 year ago

22 kB

import argparse
import glob
import os
from PIL import Image

import cv2
import math
import numpy as np
import os
import os.path as osp
import random
import time
import torch
from tqdm import tqdm

from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels
from basicsr.data.transforms import augment
from basicsr.utils import FileClient, get_root_logger, imfrombytes, img2tensor
from basicsr.utils.registry import DATASET_REGISTRY
from torch.utils import data as data
from torchvision.transforms.functional import center_crop
import torchvision.transforms as T
from torchvision.utils import save_image

from basicsr.data.degradations import random_add_gaussian_noise_pt, random_add_poisson_noise_pt
from basicsr.data.transforms import paired_random_crop
from basicsr.utils import DiffJPEG, USMSharp
from basicsr.utils.img_process_util import filter2D
from basicsr.utils.registry import MODEL_REGISTRY
from collections import OrderedDict
from torch.nn import functional as F

cfg = {
    # dataset info.
    "name": "DF2K+OST",
    "type": "RealESRGANDataset",
    "dataroot_gt": "/home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution",
    "meta_train": [
                "DIV2K/metas/DIV2K_train_HR.list",
                "Flickr2K/metas/Flickr2K.list",
                "OST/metas/OST.list",
    ],               
    "meta_test": ["DIV2K/metas/DIV2K_valid_HR.list"],

    # the first degradation process
    "resize_prob": [0.2, 0.7, 0.1],  # up, down, keep
    "resize_range": [0.15, 1.5],
    "gaussian_noise_prob": 0.5,
    "noise_range": [1, 30],
    "poisson_scale_range": [0.05, 3],
    "gray_noise_prob": 0.4,
    "jpeg_range": [30, 95],

    "blur_kernel_size": 21,
    "kernel_list": ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso'],
    "kernel_prob": [0.45, 0.25, 0.12, 0.03, 0.12, 0.03],
    "sinc_prob": 0.1,
    "blur_sigma": [0.2, 3],
    "betag_range": [0.5, 4],
    "betap_range": [1, 2],

    # the second degradation process
    "second_blur_prob": 0.8,
    "resize_prob2": [0.3, 0.4, 0.3],  # up, down, keep
    "resize_range2": [0.3, 1.2],
    "gaussian_noise_prob2": 0.5,
    "noise_range2": [1, 25],
    "poisson_scale_range2": [0.05, 2.5],
    "gray_noise_prob2": 0.4,
    "jpeg_range2": [30, 95],

    "blur_kernel_size2": 21,
    "kernel_list2": ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso'],
    "kernel_prob2": [0.45, 0.25, 0.12, 0.03, 0.12, 0.03],
    "sinc_prob2": 0.1,
    "blur_sigma2": [0.2, 1.5],
    "betag_range2": [0.5, 4],
    "betap_range2": [1, 2],

    "final_sinc_prob": 0.8,

    "gt_size": 512,
    "keep_ratio": True,
    "use_hflip": True,
    "use_rot": False,

    # data loader
    "use_shuffle": True,
    "num_worker_per_gpu": 5,
    "batch_size_per_gpu": 12,
    "dataset_enlarge_ratio": 1,
    "prefetch_mode": None,
}

def set_seed(seed=42):
    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)

    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.benchmark = False

@DATASET_REGISTRY.register()
class RealESRGANDataset(data.Dataset):
    """Dataset used for Real-ESRGAN model:
    Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data.
    It loads gt (Ground-Truth) images, and augments them.
    It also generates blur kernels and sinc kernels for generating low-quality images.
    Note that the low-quality images are processed in tensors on GPUS for faster processing.
    Args:
        opt (dict): Config for train datasets. It contains the following keys:
            dataroot_gt (str): Data root path for gt.
            meta_info (str): Path for meta information file.
            io_backend (dict): IO backend type and other kwarg.
            use_hflip (bool): Use horizontal flips.
            use_rot (bool): Use rotation (use vertical flip and transposing h and w for implementation).
            Please see more options in the codes.
    """

    def __init__(self, opt, train=True):
        super(RealESRGANDataset, self).__init__()
        self.opt = opt
        self.file_client = None

        # kernel define
        self.data_rt = opt['dataroot_gt']

        # dataload
        self.train = train
        if self.train:
            self.metas = opt['meta_train']
        else:
            self.metas = opt['meta_test']

        self.paths = []
        for meta in self.metas:
            with open(os.path.join(self.data_rt, meta)) as fin:
                self.paths += [line.strip().split(' ')[1] for line in fin]

        # Hyperparameter of Degradation
        # blur settings for the first degradation
        self.blur_kernel_size = opt['blur_kernel_size']
        self.kernel_list = opt['kernel_list']
        self.kernel_prob = opt['kernel_prob']  # a list for each kernel probability
        self.blur_sigma = opt['blur_sigma']
        self.betag_range = opt['betag_range']  # betag used in generalized Gaussian blur kernels
        self.betap_range = opt['betap_range']  # betap used in plateau blur kernels
        self.sinc_prob = opt['sinc_prob']  # the probability for sinc filters

        # blur settings for the second degradation
        self.blur_kernel_size2 = opt['blur_kernel_size2']
        self.kernel_list2 = opt['kernel_list2']
        self.kernel_prob2 = opt['kernel_prob2']
        self.blur_sigma2 = opt['blur_sigma2']
        self.betag_range2 = opt['betag_range2']
        self.betap_range2 = opt['betap_range2']
        self.sinc_prob2 = opt['sinc_prob2']

        # a final sinc filter
        self.final_sinc_prob = opt['final_sinc_prob']

        self.kernel_range = [2 * v + 1 for v in range(3, 11)]  # kernel size ranges from 7 to 21
        # TODO: kernel range is now hard-coded, should be in the configure file
        self.pulse_tensor = torch.zeros(21, 21).float()  # convolving with pulse tensor brings no blurry effect
        self.pulse_tensor[10, 10] = 1

        self.device = torch.cuda.current_device()
        self.jpeger = DiffJPEG(differentiable=False).to(self.device)  # simulate JPEG compression artifacts
        self.usm_sharpener = USMSharp().to(self.device)  # do usm sharpening
        self.resize = opt['gt_size']
        self.keep_ratio = opt['keep_ratio']

        # function
        self.crop = T.RandomCrop((self.resize, self.resize))
        self.flip = T.RandomHorizontalFlip()
        self.transform = T.Compose(
            [
                # T.ToDtype(torch.float32, scale=True),  # only support for torch 2.++
                T.ToTensor(),
            ]
        )

    def __getitem__(self, index):
        # -------------------------------- Load gt images -------------------------------- #
        gt_path = self.paths[index]
        img_gt = Image.open(gt_path).convert("RGB")

        # -------------------------------- Image Process --------------------------------
        # resize
        h, w = img_gt.height, img_gt.width
        if self.keep_ratio:
            ratio = self.resize / min(h, w)
            h_new, w_new = round(h * ratio * 1.2), round(w * ratio * 1.2)
            img_gt = img_gt.resize((w_new, h_new), resample=Image.LANCZOS)
        else:
            img_gt = img_gt.resize((self.resize, self.resize), resample=Image.LANCZOS)
        # crop and 
        img_gt = self.crop(img_gt)
        # flip (only for train)
        if self.train:
            img_gt = self.flip(img_gt)
        # transform to tensor
        img_gt = self.transform(img_gt).to(torch.float32)

        # -------------------------------- Generate Kernels --------------------------------
        kernel, kernel2, sinc_kernel = self.generate_kernel()

        # ------------------------- Generate Low Resolutino Sample -------------------------
        lq, hq = self.generate_lr({
                "gt": img_gt.unsqueeze(0),
                "kernel1": kernel,
                "kernel2": kernel2,
                "sinc_kernel": sinc_kernel,
            })

        return lq, hq, gt_path

    def generate_kernel(self, ):
        # ------------------------ Generate kernels (used in the first degradation) ------------------------ #
        kernel_size = random.choice(self.kernel_range)
        if np.random.uniform() < self.opt['sinc_prob']:
            # this sinc filter setting is for kernels ranging from [7, 21]
            if kernel_size < 13:
                omega_c = np.random.uniform(np.pi / 3, np.pi)
            else:
                omega_c = np.random.uniform(np.pi / 5, np.pi)
            kernel = circular_lowpass_kernel(omega_c, kernel_size, pad_to=False)
        else:
            kernel = random_mixed_kernels(
                self.kernel_list,
                self.kernel_prob,
                kernel_size,
                self.blur_sigma,
                self.blur_sigma, [-math.pi, math.pi],
                self.betag_range,
                self.betap_range,
                noise_range=None)
        # pad kernel
        pad_size = (21 - kernel_size) // 2
        kernel = np.pad(kernel, ((pad_size, pad_size), (pad_size, pad_size)))
        kernel = torch.FloatTensor(kernel)

        # ------------------------ Generate kernels (used in the second degradation) ------------------------ #
        kernel_size = random.choice(self.kernel_range)
        if np.random.uniform() < self.opt['sinc_prob2']:
            if kernel_size < 13:
                omega_c = np.random.uniform(np.pi / 3, np.pi)
            else:
                omega_c = np.random.uniform(np.pi / 5, np.pi)
            kernel2 = circular_lowpass_kernel(omega_c, kernel_size, pad_to=False)
        else:
            kernel2 = random_mixed_kernels(
                self.kernel_list2,
                self.kernel_prob2,
                kernel_size,
                self.blur_sigma2,
                self.blur_sigma2, [-math.pi, math.pi],
                self.betag_range2,
                self.betap_range2,
                noise_range=None)

        # pad kernel
        pad_size = (21 - kernel_size) // 2
        kernel2 = np.pad(kernel2, ((pad_size, pad_size), (pad_size, pad_size)))
        kernel2 = torch.FloatTensor(kernel2)
        
        # ------------------------------------- the final sinc kernel ------------------------------------- #
        if np.random.uniform() < self.opt['final_sinc_prob']:
            kernel_size = random.choice(self.kernel_range)
            omega_c = np.random.uniform(np.pi / 3, np.pi)
            sinc_kernel = circular_lowpass_kernel(omega_c, kernel_size, pad_to=21)
            sinc_kernel = torch.FloatTensor(sinc_kernel)
        else:
            sinc_kernel = self.pulse_tensor
        return kernel, kernel2, sinc_kernel

    def generate_lr(self, data):
        """Accept data from dataloader, and then add two-order degradations to obtain LQ images.
        """
        # training data synthesis
        self.gt = data['gt'].to(self.device)
        self.gt_usm = self.usm_sharpener(self.gt)

        self.kernel1 = data['kernel1'].to(self.device)
        self.kernel2 = data['kernel2'].to(self.device)
        self.sinc_kernel = data['sinc_kernel'].to(self.device)

        ori_h, ori_w = self.gt.size()[2:4]

        # ----------------------- The first degradation process ----------------------- #
        # blur
        out = filter2D(self.gt_usm, self.kernel1)
        # random resize
        updown_type = random.choices(['up', 'down', 'keep'], self.opt['resize_prob'])[0]
        if updown_type == 'up':
            scale = np.random.uniform(1, self.opt['resize_range'][1])
        elif updown_type == 'down':
            scale = np.random.uniform(self.opt['resize_range'][0], 1)
        else:
            scale = 1
        mode = random.choice(['area', 'bilinear', 'bicubic'])
        out = F.interpolate(out, scale_factor=scale, mode=mode)
        # add noise
        gray_noise_prob = self.opt['gray_noise_prob']
        if np.random.uniform() < self.opt['gaussian_noise_prob']:
            out = random_add_gaussian_noise_pt(
                out, sigma_range=self.opt['noise_range'], clip=True, rounds=False, gray_prob=gray_noise_prob)
        else:
            out = random_add_poisson_noise_pt(
                out,
                scale_range=self.opt['poisson_scale_range'],
                gray_prob=gray_noise_prob,
                clip=True,
                rounds=False)
        # JPEG compression
        jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range'])
        out = torch.clamp(out, 0, 1)  # clamp to [0, 1], otherwise JPEGer will result in unpleasant artifacts
        out = self.jpeger(out, quality=jpeg_p)

        # ----------------------- The second degradation process ----------------------- #
        # blur
        if np.random.uniform() < self.opt['second_blur_prob']:
            out = filter2D(out, self.kernel2)
        # random resize
        updown_type = random.choices(['up', 'down', 'keep'], self.opt['resize_prob2'])[0]
        if updown_type == 'up':
            scale = np.random.uniform(1, self.opt['resize_range2'][1])
        elif updown_type == 'down':
            scale = np.random.uniform(self.opt['resize_range2'][0], 1)
        else:
            scale = 1
        mode = random.choice(['area', 'bilinear', 'bicubic'])
        out = F.interpolate(
            out, size=(int(ori_h * scale), int(ori_w * scale)), mode=mode)
        # add noise
        gray_noise_prob = self.opt['gray_noise_prob2']
        if np.random.uniform() < self.opt['gaussian_noise_prob2']:
            out = random_add_gaussian_noise_pt(
                out, sigma_range=self.opt['noise_range2'], clip=True, rounds=False, gray_prob=gray_noise_prob)
        else:
            out = random_add_poisson_noise_pt(
                out,
                scale_range=self.opt['poisson_scale_range2'],
                gray_prob=gray_noise_prob,
                clip=True,
                rounds=False)

        # JPEG compression + the final sinc filter
        # We also need to resize images to desired sizes. We group [resize back + sinc filter] together
        # as one operation.
        # We consider two orders:
        #   1. [resize back + sinc filter] + JPEG compression
        #   2. JPEG compression + [resize back + sinc filter]
        # Empirically, we find other combinations (sinc + JPEG + Resize) will introduce twisted lines.
        if np.random.uniform() < 0.5:
            # resize back + the final sinc filter
            mode = random.choice(['area', 'bilinear', 'bicubic'])
            out = F.interpolate(out, size=(ori_h, ori_w), mode=mode)
            out = filter2D(out, self.sinc_kernel)
            # JPEG compression
            jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range2'])
            out = torch.clamp(out, 0, 1)
            out = self.jpeger(out, quality=jpeg_p)
        else:
            # JPEG compression
            jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range2'])
            out = torch.clamp(out, 0, 1)
            out = self.jpeger(out, quality=jpeg_p)
            # resize back + the final sinc filter
            mode = random.choice(['area', 'bilinear', 'bicubic'])
            out = F.interpolate(out, size=(ori_h, ori_w), mode=mode)
            out = filter2D(out, self.sinc_kernel)

        # clamp and round
        lq = torch.clamp((out * 255.0).round(), 0, 255) / 255.
        lq = lq.contiguous()  # for the warning: grad and param do not obey the gradient layout contract

        hq = self.usm_sharpener(self.gt)
        return lq[0], hq[0] 

    def __len__(self):
        return len(self.paths)

def real_esrgan_sampler():
    """
    Generate multi-scale versions for GT images with LANCZOS resampling.
    It is now used for DF2K dataset (DIV2K + Flickr 2K)
    """
    parser = argparse.ArgumentParser()
    parser.add_argument('--num_samples', type=int, default=3, help='train: one to many')
    args = parser.parse_args()

    # generate training dataset
    dataset = RealESRGANDataset(cfg, train=True)
    data_dl = data.DataLoader( 
        dataset,
        batch_size = 1
    )
    print("Train Data:", dataset.data_rt, len(data_dl))
    for number in range(args.num_samples):
        print("="*100)
        print(f"Generate round {number}...")

        meta_info = {}
        for sample in tqdm(data_dl):
            lq, hq, path = sample
            # /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/0098.png
            file_name = os.path.basename(path[0])
            gt_folder = os.path.dirname(path[0]) # /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/
            if "DIV2K_train_HR" in gt_folder or "DIV2K_valid_HR" in gt_folder:
                hq_folder = gt_folder.replace("HR", f"pair/SR{number+1}/HR")
                lq_folder = gt_folder.replace("HR", f"pair/SR{number+1}/LR")
            else:
                hq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR{number+1}"), "HR/")
                lq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR{number+1}"), "LR/")

            os.makedirs(hq_folder, exist_ok=True)
            os.makedirs(lq_folder, exist_ok=True)

            hq_path = os.path.join(hq_folder, file_name)
            lq_path = os.path.join(lq_folder, file_name)

            save_image(hq[0], hq_path)
            save_image(lq[0], lq_path)

            dset = os.path.relpath(gt_folder, dataset.data_rt).split("/")[0]
            if dset not in meta_info:
                meta_info[dset] = [(lq_path, hq_path)]
            else:
                meta_info[dset].append((lq_path, hq_path))

        for dset, dlist in meta_info.items():
            with open(os.path.join(dataset.data_rt,'{}/metas/{}_train_SR{}.list'.format(dset, dset, number+1)), 'w') as fp:
                for item in dlist:
                    fp.write('{} {} {}\n'.format(item[0], item[1], None))
            print(os.path.join(dataset.data_rt,'{}/metas/{}_train_SR{}.list'.format(dset, dset, number+1)), len(dlist))


    # generate testing dataset
    dataset = RealESRGANDataset(cfg, train=False)
    data_dl = data.DataLoader( 
        dataset,
        batch_size = 1
    )
    print("Test Data:", dataset.data_rt, len(data_dl))
    print("="*100)
    print(f"Generate round {number}...")

    meta_info = {}
    for sample in tqdm(data_dl):
        lq, hq, path = sample
        # /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/0098.png
        file_name = os.path.basename(path[0])
        gt_folder = os.path.dirname(path[0]) # /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/
        if "DIV2K_train_HR" in gt_folder or "DIV2K_valid_HR" in gt_folder:
            hq_folder = gt_folder.replace("HR", f"pair/SR/HR")
            lq_folder = gt_folder.replace("HR", f"pair/SR/LR")
        else:
            hq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR"), "HR/")
            lq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR"), "LR/")

        os.makedirs(hq_folder, exist_ok=True)
        os.makedirs(lq_folder, exist_ok=True)

        hq_path = os.path.join(hq_folder, file_name)
        lq_path = os.path.join(lq_folder, file_name)

        save_image(hq[0], hq_path)
        save_image(lq[0], lq_path)

        dset = os.path.relpath(gt_folder, dataset.data_rt).split("/")[0]
        if dset not in meta_info:
            meta_info[dset] = [(lq_path, hq_path)]
        else:
            meta_info[dset].append((lq_path, hq_path))

    for dset, dlist in meta_info.items():
        with open(os.path.join(dataset.data_rt,'{}/metas/{}_valid_SR.list'.format(dset, dset)), 'w') as fp:
            for item in dlist:
                fp.write('{} {} {}\n'.format(item[0], item[1], None))
        print(os.path.join(dataset.data_rt,'{}/metas/{}_valid_SR.list'.format(dset, dset)), len(dlist))

def simple_multiscale():
    """
    Generate multi-scale versions for GT images with LANCZOS resampling.
    It is now used for DF2K dataset (DIV2K + Flickr 2K)
    """
    parser = argparse.ArgumentParser()
    parser.add_argument('--input', type=str, default='DIV2K/DIV2K_train_HR', help='Input folder')
    parser.add_argument('--output', type=str, default='DIV2K/DIV2K_train_multiscale', help='Output folder')
    args = parser.parse_args()
    os.makedirs(args.output, exist_ok=True)

    # For DF2K, we consider the following three scales,
    # and the smallest image whose shortest edge is 400
    scale_list = [0.75, 0.5, 1 / 3]
    shortest_edge = 400

    path_list = sorted(glob.glob(os.path.join(args.input, '*')))
    for path in path_list:
        basename = os.path.splitext(os.path.basename(path))[0]

        img = Image.open(path)
        width, height = img.size
        for idx, scale in enumerate(scale_list):
            print(f'\t{scale:.2f}')
            rlt = img.resize((int(width * scale), int(height * scale)), resample=Image.LANCZOS)
            rlt = rlt.resize((width, height), resample=Image.NEAREST)
            rlt.save(os.path.join(args.output, f'{basename}T{idx}.png'))

        # save the smallest image which the shortest edge is 400
        if width < height:
            ratio = height / width
            width = shortest_edge
            height = int(width * ratio)
        else:
            ratio = width / height
            height = shortest_edge
            width = int(height * ratio)
        rlt = img.resize((int(width), int(height)), resample=Image.LANCZOS)
        rlt = rlt.resize(img.size, resample=Image.NEAREST)
        rlt.save(os.path.join(args.output, f'{basename}T{idx+1}.png'))


if __name__ == '__main__':
    set_seed(1229)
    # simple version
    # simple_multiscale()
    
    # Real-ESRGAN for data generation
    real_esrgan_sampler()

# python 2_generate_lowresolution.py --num_samples 3

	import argparse
	import glob
	import os
	from PIL import Image

	import cv2
	import math
	import numpy as np
	import os
	import os.path as osp
	import random
	import time
	import torch
	from tqdm import tqdm

	from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels
	from basicsr.data.transforms import augment
	from basicsr.utils import FileClient, get_root_logger, imfrombytes, img2tensor
	from basicsr.utils.registry import DATASET_REGISTRY
	from torch.utils import data as data
	from torchvision.transforms.functional import center_crop
	import torchvision.transforms as T
	from torchvision.utils import save_image

	from basicsr.data.degradations import random_add_gaussian_noise_pt, random_add_poisson_noise_pt
	from basicsr.data.transforms import paired_random_crop
	from basicsr.utils import DiffJPEG, USMSharp
	from basicsr.utils.img_process_util import filter2D
	from basicsr.utils.registry import MODEL_REGISTRY
	from collections import OrderedDict
	from torch.nn import functional as F

	cfg = {
	# dataset info.
	"name": "DF2K+OST",
	"type": "RealESRGANDataset",
	"dataroot_gt": "/home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution",
	"meta_train": [
	"DIV2K/metas/DIV2K_train_HR.list",
	"Flickr2K/metas/Flickr2K.list",
	"OST/metas/OST.list",
	],
	"meta_test": ["DIV2K/metas/DIV2K_valid_HR.list"],

	# the first degradation process
	"resize_prob": [0.2, 0.7, 0.1], # up, down, keep
	"resize_range": [0.15, 1.5],
	"gaussian_noise_prob": 0.5,
	"noise_range": [1, 30],
	"poisson_scale_range": [0.05, 3],
	"gray_noise_prob": 0.4,
	"jpeg_range": [30, 95],

	"blur_kernel_size": 21,
	"kernel_list": ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso'],
	"kernel_prob": [0.45, 0.25, 0.12, 0.03, 0.12, 0.03],
	"sinc_prob": 0.1,
	"blur_sigma": [0.2, 3],
	"betag_range": [0.5, 4],
	"betap_range": [1, 2],

	# the second degradation process
	"second_blur_prob": 0.8,
	"resize_prob2": [0.3, 0.4, 0.3], # up, down, keep
	"resize_range2": [0.3, 1.2],
	"gaussian_noise_prob2": 0.5,
	"noise_range2": [1, 25],
	"poisson_scale_range2": [0.05, 2.5],
	"gray_noise_prob2": 0.4,
	"jpeg_range2": [30, 95],

	"blur_kernel_size2": 21,
	"kernel_list2": ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso'],
	"kernel_prob2": [0.45, 0.25, 0.12, 0.03, 0.12, 0.03],
	"sinc_prob2": 0.1,
	"blur_sigma2": [0.2, 1.5],
	"betag_range2": [0.5, 4],
	"betap_range2": [1, 2],

	"final_sinc_prob": 0.8,

	"gt_size": 512,
	"keep_ratio": True,
	"use_hflip": True,
	"use_rot": False,

	# data loader
	"use_shuffle": True,
	"num_worker_per_gpu": 5,
	"batch_size_per_gpu": 12,
	"dataset_enlarge_ratio": 1,
	"prefetch_mode": None,
	}

	def set_seed(seed=42):
	random.seed(seed)
	np.random.seed(seed)
	torch.manual_seed(seed)
	torch.cuda.manual_seed(seed)
	torch.cuda.manual_seed_all(seed)

	torch.backends.cudnn.deterministic = True
	torch.backends.cudnn.benchmark = False

	@DATASET_REGISTRY.register()
	class RealESRGANDataset(data.Dataset):
	"""Dataset used for Real-ESRGAN model:
	Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data.

	It loads gt (Ground-Truth) images, and augments them.
	It also generates blur kernels and sinc kernels for generating low-quality images.
	Note that the low-quality images are processed in tensors on GPUS for faster processing.

	Args:
	opt (dict): Config for train datasets. It contains the following keys:
	dataroot_gt (str): Data root path for gt.
	meta_info (str): Path for meta information file.
	io_backend (dict): IO backend type and other kwarg.
	use_hflip (bool): Use horizontal flips.
	use_rot (bool): Use rotation (use vertical flip and transposing h and w for implementation).
	Please see more options in the codes.
	"""

	def __init__(self, opt, train=True):
	super(RealESRGANDataset, self).__init__()
	self.opt = opt
	self.file_client = None

	# kernel define
	self.data_rt = opt['dataroot_gt']

	# dataload
	self.train = train
	if self.train:
	self.metas = opt['meta_train']
	else:
	self.metas = opt['meta_test']

	self.paths = []
	for meta in self.metas:
	with open(os.path.join(self.data_rt, meta)) as fin:
	self.paths += [line.strip().split(' ')[1] for line in fin]

	# Hyperparameter of Degradation
	# blur settings for the first degradation
	self.blur_kernel_size = opt['blur_kernel_size']
	self.kernel_list = opt['kernel_list']
	self.kernel_prob = opt['kernel_prob'] # a list for each kernel probability
	self.blur_sigma = opt['blur_sigma']
	self.betag_range = opt['betag_range'] # betag used in generalized Gaussian blur kernels
	self.betap_range = opt['betap_range'] # betap used in plateau blur kernels
	self.sinc_prob = opt['sinc_prob'] # the probability for sinc filters

	# blur settings for the second degradation
	self.blur_kernel_size2 = opt['blur_kernel_size2']
	self.kernel_list2 = opt['kernel_list2']
	self.kernel_prob2 = opt['kernel_prob2']
	self.blur_sigma2 = opt['blur_sigma2']
	self.betag_range2 = opt['betag_range2']
	self.betap_range2 = opt['betap_range2']
	self.sinc_prob2 = opt['sinc_prob2']

	# a final sinc filter
	self.final_sinc_prob = opt['final_sinc_prob']

	self.kernel_range = [2 * v + 1 for v in range(3, 11)] # kernel size ranges from 7 to 21
	# TODO: kernel range is now hard-coded, should be in the configure file
	self.pulse_tensor = torch.zeros(21, 21).float() # convolving with pulse tensor brings no blurry effect
	self.pulse_tensor[10, 10] = 1

	self.device = torch.cuda.current_device()
	self.jpeger = DiffJPEG(differentiable=False).to(self.device) # simulate JPEG compression artifacts
	self.usm_sharpener = USMSharp().to(self.device) # do usm sharpening
	self.resize = opt['gt_size']
	self.keep_ratio = opt['keep_ratio']

	# function
	self.crop = T.RandomCrop((self.resize, self.resize))
	self.flip = T.RandomHorizontalFlip()
	self.transform = T.Compose(
	[
	# T.ToDtype(torch.float32, scale=True), # only support for torch 2.++
	T.ToTensor(),
	]
	)

	def __getitem__(self, index):
	# -------------------------------- Load gt images -------------------------------- #
	gt_path = self.paths[index]
	img_gt = Image.open(gt_path).convert("RGB")

	# -------------------------------- Image Process --------------------------------
	# resize
	h, w = img_gt.height, img_gt.width
	if self.keep_ratio:
	ratio = self.resize / min(h, w)
	h_new, w_new = round(h * ratio * 1.2), round(w * ratio * 1.2)
	img_gt = img_gt.resize((w_new, h_new), resample=Image.LANCZOS)
	else:
	img_gt = img_gt.resize((self.resize, self.resize), resample=Image.LANCZOS)
	# crop and
	img_gt = self.crop(img_gt)
	# flip (only for train)
	if self.train:
	img_gt = self.flip(img_gt)
	# transform to tensor
	img_gt = self.transform(img_gt).to(torch.float32)

	# -------------------------------- Generate Kernels --------------------------------
	kernel, kernel2, sinc_kernel = self.generate_kernel()

	# ------------------------- Generate Low Resolutino Sample -------------------------
	lq, hq = self.generate_lr({
	"gt": img_gt.unsqueeze(0),
	"kernel1": kernel,
	"kernel2": kernel2,
	"sinc_kernel": sinc_kernel,
	})

	return lq, hq, gt_path

	def generate_kernel(self, ):
	# ------------------------ Generate kernels (used in the first degradation) ------------------------ #
	kernel_size = random.choice(self.kernel_range)
	if np.random.uniform() < self.opt['sinc_prob']:
	# this sinc filter setting is for kernels ranging from [7, 21]
	if kernel_size < 13:
	omega_c = np.random.uniform(np.pi / 3, np.pi)
	else:
	omega_c = np.random.uniform(np.pi / 5, np.pi)
	kernel = circular_lowpass_kernel(omega_c, kernel_size, pad_to=False)
	else:
	kernel = random_mixed_kernels(
	self.kernel_list,
	self.kernel_prob,
	kernel_size,
	self.blur_sigma,
	self.blur_sigma, [-math.pi, math.pi],
	self.betag_range,
	self.betap_range,
	noise_range=None)
	# pad kernel
	pad_size = (21 - kernel_size) // 2
	kernel = np.pad(kernel, ((pad_size, pad_size), (pad_size, pad_size)))
	kernel = torch.FloatTensor(kernel)

	# ------------------------ Generate kernels (used in the second degradation) ------------------------ #
	kernel_size = random.choice(self.kernel_range)
	if np.random.uniform() < self.opt['sinc_prob2']:
	if kernel_size < 13:
	omega_c = np.random.uniform(np.pi / 3, np.pi)
	else:
	omega_c = np.random.uniform(np.pi / 5, np.pi)
	kernel2 = circular_lowpass_kernel(omega_c, kernel_size, pad_to=False)
	else:
	kernel2 = random_mixed_kernels(
	self.kernel_list2,
	self.kernel_prob2,
	kernel_size,
	self.blur_sigma2,
	self.blur_sigma2, [-math.pi, math.pi],
	self.betag_range2,
	self.betap_range2,
	noise_range=None)

	# pad kernel
	pad_size = (21 - kernel_size) // 2
	kernel2 = np.pad(kernel2, ((pad_size, pad_size), (pad_size, pad_size)))
	kernel2 = torch.FloatTensor(kernel2)

	# ------------------------------------- the final sinc kernel ------------------------------------- #
	if np.random.uniform() < self.opt['final_sinc_prob']:
	kernel_size = random.choice(self.kernel_range)
	omega_c = np.random.uniform(np.pi / 3, np.pi)
	sinc_kernel = circular_lowpass_kernel(omega_c, kernel_size, pad_to=21)
	sinc_kernel = torch.FloatTensor(sinc_kernel)
	else:
	sinc_kernel = self.pulse_tensor
	return kernel, kernel2, sinc_kernel

	def generate_lr(self, data):
	"""Accept data from dataloader, and then add two-order degradations to obtain LQ images.
	"""
	# training data synthesis
	self.gt = data['gt'].to(self.device)
	self.gt_usm = self.usm_sharpener(self.gt)

	self.kernel1 = data['kernel1'].to(self.device)
	self.kernel2 = data['kernel2'].to(self.device)
	self.sinc_kernel = data['sinc_kernel'].to(self.device)

	ori_h, ori_w = self.gt.size()[2:4]

	# ----------------------- The first degradation process ----------------------- #
	# blur
	out = filter2D(self.gt_usm, self.kernel1)
	# random resize
	updown_type = random.choices(['up', 'down', 'keep'], self.opt['resize_prob'])[0]
	if updown_type == 'up':
	scale = np.random.uniform(1, self.opt['resize_range'][1])
	elif updown_type == 'down':
	scale = np.random.uniform(self.opt['resize_range'][0], 1)
	else:
	scale = 1
	mode = random.choice(['area', 'bilinear', 'bicubic'])
	out = F.interpolate(out, scale_factor=scale, mode=mode)
	# add noise
	gray_noise_prob = self.opt['gray_noise_prob']
	if np.random.uniform() < self.opt['gaussian_noise_prob']:
	out = random_add_gaussian_noise_pt(
	out, sigma_range=self.opt['noise_range'], clip=True, rounds=False, gray_prob=gray_noise_prob)
	else:
	out = random_add_poisson_noise_pt(
	out,
	scale_range=self.opt['poisson_scale_range'],
	gray_prob=gray_noise_prob,
	clip=True,
	rounds=False)
	# JPEG compression
	jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range'])
	out = torch.clamp(out, 0, 1) # clamp to [0, 1], otherwise JPEGer will result in unpleasant artifacts
	out = self.jpeger(out, quality=jpeg_p)

	# ----------------------- The second degradation process ----------------------- #
	# blur
	if np.random.uniform() < self.opt['second_blur_prob']:
	out = filter2D(out, self.kernel2)
	# random resize
	updown_type = random.choices(['up', 'down', 'keep'], self.opt['resize_prob2'])[0]
	if updown_type == 'up':
	scale = np.random.uniform(1, self.opt['resize_range2'][1])
	elif updown_type == 'down':
	scale = np.random.uniform(self.opt['resize_range2'][0], 1)
	else:
	scale = 1
	mode = random.choice(['area', 'bilinear', 'bicubic'])
	out = F.interpolate(
	out, size=(int(ori_h * scale), int(ori_w * scale)), mode=mode)
	# add noise
	gray_noise_prob = self.opt['gray_noise_prob2']
	if np.random.uniform() < self.opt['gaussian_noise_prob2']:
	out = random_add_gaussian_noise_pt(
	out, sigma_range=self.opt['noise_range2'], clip=True, rounds=False, gray_prob=gray_noise_prob)
	else:
	out = random_add_poisson_noise_pt(
	out,
	scale_range=self.opt['poisson_scale_range2'],
	gray_prob=gray_noise_prob,
	clip=True,
	rounds=False)

	# JPEG compression + the final sinc filter
	# We also need to resize images to desired sizes. We group [resize back + sinc filter] together
	# as one operation.
	# We consider two orders:
	# 1. [resize back + sinc filter] + JPEG compression
	# 2. JPEG compression + [resize back + sinc filter]
	# Empirically, we find other combinations (sinc + JPEG + Resize) will introduce twisted lines.
	if np.random.uniform() < 0.5:
	# resize back + the final sinc filter
	mode = random.choice(['area', 'bilinear', 'bicubic'])
	out = F.interpolate(out, size=(ori_h, ori_w), mode=mode)
	out = filter2D(out, self.sinc_kernel)
	# JPEG compression
	jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range2'])
	out = torch.clamp(out, 0, 1)
	out = self.jpeger(out, quality=jpeg_p)
	else:
	# JPEG compression
	jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range2'])
	out = torch.clamp(out, 0, 1)
	out = self.jpeger(out, quality=jpeg_p)
	# resize back + the final sinc filter
	mode = random.choice(['area', 'bilinear', 'bicubic'])
	out = F.interpolate(out, size=(ori_h, ori_w), mode=mode)
	out = filter2D(out, self.sinc_kernel)

	# clamp and round
	lq = torch.clamp((out * 255.0).round(), 0, 255) / 255.
	lq = lq.contiguous() # for the warning: grad and param do not obey the gradient layout contract

	hq = self.usm_sharpener(self.gt)
	return lq[0], hq[0]

	def __len__(self):
	return len(self.paths)

	def real_esrgan_sampler():
	"""
	Generate multi-scale versions for GT images with LANCZOS resampling.
	It is now used for DF2K dataset (DIV2K + Flickr 2K)
	"""
	parser = argparse.ArgumentParser()
	parser.add_argument('--num_samples', type=int, default=3, help='train: one to many')
	args = parser.parse_args()

	# generate training dataset
	dataset = RealESRGANDataset(cfg, train=True)
	data_dl = data.DataLoader(
	dataset,
	batch_size = 1
	)
	print("Train Data:", dataset.data_rt, len(data_dl))
	for number in range(args.num_samples):
	print("="*100)
	print(f"Generate round {number}...")

	meta_info = {}
	for sample in tqdm(data_dl):
	lq, hq, path = sample
	# /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/0098.png
	file_name = os.path.basename(path[0])
	gt_folder = os.path.dirname(path[0]) # /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/
	if "DIV2K_train_HR" in gt_folder or "DIV2K_valid_HR" in gt_folder:
	hq_folder = gt_folder.replace("HR", f"pair/SR{number+1}/HR")
	lq_folder = gt_folder.replace("HR", f"pair/SR{number+1}/LR")
	else:
	hq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR{number+1}"), "HR/")
	lq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR{number+1}"), "LR/")

	os.makedirs(hq_folder, exist_ok=True)
	os.makedirs(lq_folder, exist_ok=True)

	hq_path = os.path.join(hq_folder, file_name)
	lq_path = os.path.join(lq_folder, file_name)

	save_image(hq[0], hq_path)
	save_image(lq[0], lq_path)

	dset = os.path.relpath(gt_folder, dataset.data_rt).split("/")[0]
	if dset not in meta_info:
	meta_info[dset] = [(lq_path, hq_path)]
	else:
	meta_info[dset].append((lq_path, hq_path))

	for dset, dlist in meta_info.items():
	with open(os.path.join(dataset.data_rt,'{}/metas/{}_train_SR{}.list'.format(dset, dset, number+1)), 'w') as fp:
	for item in dlist:
	fp.write('{} {} {}\n'.format(item[0], item[1], None))
	print(os.path.join(dataset.data_rt,'{}/metas/{}_train_SR{}.list'.format(dset, dset, number+1)), len(dlist))


	# generate testing dataset
	dataset = RealESRGANDataset(cfg, train=False)
	data_dl = data.DataLoader(
	dataset,
	batch_size = 1
	)
	print("Test Data:", dataset.data_rt, len(data_dl))
	print("="*100)
	print(f"Generate round {number}...")

	meta_info = {}
	for sample in tqdm(data_dl):
	lq, hq, path = sample
	# /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/0098.png
	file_name = os.path.basename(path[0])
	gt_folder = os.path.dirname(path[0]) # /home/CORP/hsiang.chen/Project/Datasets/IR/SuperResolution/DIV2K/DIV2K_train_HR/
	if "DIV2K_train_HR" in gt_folder or "DIV2K_valid_HR" in gt_folder:
	hq_folder = gt_folder.replace("HR", f"pair/SR/HR")
	lq_folder = gt_folder.replace("HR", f"pair/SR/LR")
	else:
	hq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR"), "HR/")
	lq_folder = os.path.join(gt_folder.replace("images", f"images_pair/SR"), "LR/")

	os.makedirs(hq_folder, exist_ok=True)
	os.makedirs(lq_folder, exist_ok=True)

	hq_path = os.path.join(hq_folder, file_name)
	lq_path = os.path.join(lq_folder, file_name)

	save_image(hq[0], hq_path)
	save_image(lq[0], lq_path)

	dset = os.path.relpath(gt_folder, dataset.data_rt).split("/")[0]
	if dset not in meta_info:
	meta_info[dset] = [(lq_path, hq_path)]
	else:
	meta_info[dset].append((lq_path, hq_path))

	for dset, dlist in meta_info.items():
	with open(os.path.join(dataset.data_rt,'{}/metas/{}_valid_SR.list'.format(dset, dset)), 'w') as fp:
	for item in dlist:
	fp.write('{} {} {}\n'.format(item[0], item[1], None))
	print(os.path.join(dataset.data_rt,'{}/metas/{}_valid_SR.list'.format(dset, dset)), len(dlist))

	def simple_multiscale():
	"""
	Generate multi-scale versions for GT images with LANCZOS resampling.
	It is now used for DF2K dataset (DIV2K + Flickr 2K)
	"""
	parser = argparse.ArgumentParser()
	parser.add_argument('--input', type=str, default='DIV2K/DIV2K_train_HR', help='Input folder')
	parser.add_argument('--output', type=str, default='DIV2K/DIV2K_train_multiscale', help='Output folder')
	args = parser.parse_args()
	os.makedirs(args.output, exist_ok=True)

	# For DF2K, we consider the following three scales,
	# and the smallest image whose shortest edge is 400
	scale_list = [0.75, 0.5, 1 / 3]
	shortest_edge = 400

	path_list = sorted(glob.glob(os.path.join(args.input, '*')))
	for path in path_list:
	basename = os.path.splitext(os.path.basename(path))[0]

	img = Image.open(path)
	width, height = img.size
	for idx, scale in enumerate(scale_list):
	print(f'\t{scale:.2f}')
	rlt = img.resize((int(width * scale), int(height * scale)), resample=Image.LANCZOS)
	rlt = rlt.resize((width, height), resample=Image.NEAREST)
	rlt.save(os.path.join(args.output, f'{basename}T{idx}.png'))

	# save the smallest image which the shortest edge is 400
	if width < height:
	ratio = height / width
	width = shortest_edge
	height = int(width * ratio)
	else:
	ratio = width / height
	height = shortest_edge
	width = int(height * ratio)
	rlt = img.resize((int(width), int(height)), resample=Image.LANCZOS)
	rlt = rlt.resize(img.size, resample=Image.NEAREST)
	rlt.save(os.path.join(args.output, f'{basename}T{idx+1}.png'))


	if __name__ == '__main__':
	set_seed(1229)
	# simple version
	# simple_multiscale()

	# Real-ESRGAN for data generation
	real_esrgan_sampler()

	# python 2_generate_lowresolution.py --num_samples 3