|
import numpy as np |
|
import random |
|
import torch |
|
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.models.srgan_model import SRGANModel |
|
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 |
|
|
|
|
|
@MODEL_REGISTRY.register() |
|
class RealESRGANModel(SRGANModel): |
|
"""RealESRGAN Model for Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data. |
|
|
|
It mainly performs: |
|
1. randomly synthesize LQ images in GPU tensors |
|
2. optimize the networks with GAN training. |
|
""" |
|
|
|
def __init__(self, opt): |
|
super(RealESRGANModel, self).__init__(opt) |
|
self.jpeger = DiffJPEG(differentiable=False).cuda() |
|
self.usm_sharpener = USMSharp().cuda() |
|
self.queue_size = opt.get('queue_size', 180) |
|
|
|
@torch.no_grad() |
|
def _dequeue_and_enqueue(self): |
|
"""It is the training pair pool for increasing the diversity in a batch. |
|
|
|
Batch processing limits the diversity of synthetic degradations in a batch. For example, samples in a |
|
batch could not have different resize scaling factors. Therefore, we employ this training pair pool |
|
to increase the degradation diversity in a batch. |
|
""" |
|
|
|
b, c, h, w = self.lq.size() |
|
if not hasattr(self, 'queue_lr'): |
|
assert self.queue_size % b == 0, f'queue size {self.queue_size} should be divisible by batch size {b}' |
|
self.queue_lr = torch.zeros(self.queue_size, c, h, w).cuda() |
|
_, c, h, w = self.gt.size() |
|
self.queue_gt = torch.zeros(self.queue_size, c, h, w).cuda() |
|
self.queue_ptr = 0 |
|
if self.queue_ptr == self.queue_size: |
|
|
|
|
|
idx = torch.randperm(self.queue_size) |
|
self.queue_lr = self.queue_lr[idx] |
|
self.queue_gt = self.queue_gt[idx] |
|
|
|
lq_dequeue = self.queue_lr[0:b, :, :, :].clone() |
|
gt_dequeue = self.queue_gt[0:b, :, :, :].clone() |
|
|
|
self.queue_lr[0:b, :, :, :] = self.lq.clone() |
|
self.queue_gt[0:b, :, :, :] = self.gt.clone() |
|
|
|
self.lq = lq_dequeue |
|
self.gt = gt_dequeue |
|
else: |
|
|
|
self.queue_lr[self.queue_ptr:self.queue_ptr + b, :, :, :] = self.lq.clone() |
|
self.queue_gt[self.queue_ptr:self.queue_ptr + b, :, :, :] = self.gt.clone() |
|
self.queue_ptr = self.queue_ptr + b |
|
|
|
@torch.no_grad() |
|
def feed_data(self, data): |
|
"""Accept data from dataloader, and then add two-order degradations to obtain LQ images. |
|
""" |
|
if self.is_train and self.opt.get('high_order_degradation', True): |
|
|
|
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] |
|
|
|
|
|
|
|
out = filter2D(self.gt_usm, self.kernel1) |
|
|
|
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) |
|
|
|
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_p = out.new_zeros(out.size(0)).uniform_(*self.opt['jpeg_range']) |
|
out = torch.clamp(out, 0, 1) |
|
out = self.jpeger(out, quality=jpeg_p) |
|
|
|
|
|
|
|
if np.random.uniform() < self.opt['second_blur_prob']: |
|
out = filter2D(out, self.kernel2) |
|
|
|
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 / self.opt['scale'] * scale), int(ori_w / self.opt['scale'] * scale)), mode=mode) |
|
|
|
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) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if np.random.uniform() < 0.5: |
|
|
|
mode = random.choice(['area', 'bilinear', 'bicubic']) |
|
out = F.interpolate(out, size=(ori_h // self.opt['scale'], ori_w // self.opt['scale']), mode=mode) |
|
out = filter2D(out, self.sinc_kernel) |
|
|
|
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_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) |
|
|
|
mode = random.choice(['area', 'bilinear', 'bicubic']) |
|
out = F.interpolate(out, size=(ori_h // self.opt['scale'], ori_w // self.opt['scale']), mode=mode) |
|
out = filter2D(out, self.sinc_kernel) |
|
|
|
|
|
self.lq = torch.clamp((out * 255.0).round(), 0, 255) / 255. |
|
|
|
|
|
gt_size = self.opt['gt_size'] |
|
(self.gt, self.gt_usm), self.lq = paired_random_crop([self.gt, self.gt_usm], self.lq, gt_size, |
|
self.opt['scale']) |
|
|
|
|
|
self._dequeue_and_enqueue() |
|
|
|
self.gt_usm = self.usm_sharpener(self.gt) |
|
self.lq = self.lq.contiguous() |
|
else: |
|
|
|
self.lq = data['lq'].to(self.device) |
|
if 'gt' in data: |
|
self.gt = data['gt'].to(self.device) |
|
self.gt_usm = self.usm_sharpener(self.gt) |
|
|
|
def nondist_validation(self, dataloader, current_iter, tb_logger, save_img): |
|
|
|
self.is_train = False |
|
super(RealESRGANModel, self).nondist_validation(dataloader, current_iter, tb_logger, save_img) |
|
self.is_train = True |
|
|
|
def optimize_parameters(self, current_iter): |
|
|
|
l1_gt = self.gt_usm |
|
percep_gt = self.gt_usm |
|
gan_gt = self.gt_usm |
|
if self.opt['l1_gt_usm'] is False: |
|
l1_gt = self.gt |
|
if self.opt['percep_gt_usm'] is False: |
|
percep_gt = self.gt |
|
if self.opt['gan_gt_usm'] is False: |
|
gan_gt = self.gt |
|
|
|
|
|
for p in self.net_d.parameters(): |
|
p.requires_grad = False |
|
|
|
self.optimizer_g.zero_grad() |
|
self.output = self.net_g(self.lq) |
|
|
|
l_g_total = 0 |
|
loss_dict = OrderedDict() |
|
if (current_iter % self.net_d_iters == 0 and current_iter > self.net_d_init_iters): |
|
|
|
if self.cri_pix: |
|
l_g_pix = self.cri_pix(self.output, l1_gt) |
|
l_g_total += l_g_pix |
|
loss_dict['l_g_pix'] = l_g_pix |
|
|
|
if self.cri_perceptual: |
|
l_g_percep, l_g_style = self.cri_perceptual(self.output, percep_gt) |
|
if l_g_percep is not None: |
|
l_g_total += l_g_percep |
|
loss_dict['l_g_percep'] = l_g_percep |
|
if l_g_style is not None: |
|
l_g_total += l_g_style |
|
loss_dict['l_g_style'] = l_g_style |
|
|
|
fake_g_pred = self.net_d(self.output) |
|
l_g_gan = self.cri_gan(fake_g_pred, True, is_disc=False) |
|
l_g_total += l_g_gan |
|
loss_dict['l_g_gan'] = l_g_gan |
|
|
|
l_g_total.backward() |
|
self.optimizer_g.step() |
|
|
|
|
|
for p in self.net_d.parameters(): |
|
p.requires_grad = True |
|
|
|
self.optimizer_d.zero_grad() |
|
|
|
real_d_pred = self.net_d(gan_gt) |
|
l_d_real = self.cri_gan(real_d_pred, True, is_disc=True) |
|
loss_dict['l_d_real'] = l_d_real |
|
loss_dict['out_d_real'] = torch.mean(real_d_pred.detach()) |
|
l_d_real.backward() |
|
|
|
fake_d_pred = self.net_d(self.output.detach().clone()) |
|
l_d_fake = self.cri_gan(fake_d_pred, False, is_disc=True) |
|
loss_dict['l_d_fake'] = l_d_fake |
|
loss_dict['out_d_fake'] = torch.mean(fake_d_pred.detach()) |
|
l_d_fake.backward() |
|
self.optimizer_d.step() |
|
|
|
if self.ema_decay > 0: |
|
self.model_ema(decay=self.ema_decay) |
|
|
|
self.log_dict = self.reduce_loss_dict(loss_dict) |
|
|