def update(self, gt, pred):
C, H, W = gt.size()[-3:]
if isinstance(gt, torch.Tensor):
gt = Variable(gt)
if isinstance(pred, torch.Tensor):
pred = Variable(pred)
mse_score = self.mse_loss(pred, gt)
eps = 1e-4
pred.data[pred.data < eps] = eps
pred.data[pred.data > 1 - eps] = 1 -eps
bce_score = self.bce_loss(pred, gt)
# https://github.com/Po-Hsun-Su/pytorch-ssim
ssim_score = pytorch_ssim.ssim(pred[:,:,0], gt[:,:,0])
psnr_score = 10 * math.log10(1 / mse_score.item())
bce_score = bce_score.item() * C * H * W
mse_score = mse_score.item() * C * H * W
self.bce_results.append(bce_score)
self.mse_results.append(mse_score)
self.psnr_results.append(psnr_score)
self.ssim_results.append(ssim_score)
def get_ssim_torch(tensor1, tensor2):
torch_ssim = ssim(tensor1, tensor2)
return torch_ssim
batch_size = val_lr.size(0)
valing_results['batch_sizes'] += batch_size
lr = val_lr.float()
hr = val_hr.float()
if torch.cuda.is_available():
lr = lr.cuda()
hr = hr.cuda()
sr = netG(lr)
if cfg.TRAIN.use_seg:
seg = netSeg(sr)[:, :cfg.DATASET.n_classes, ...].cuda()
val_seg = val_seg.cuda()
validate_seg(val_seg, seg, valing_results, batch_size,
cfg)
batch_mse = ((sr - hr)**2).data.mean()
valing_results['mse'] += batch_mse * batch_size
batch_ssim = pytorch_ssim.ssim(sr, hr).item()
valing_results['ssims'] += batch_ssim * batch_size
valing_results['psnr'] = 10 * log10(
(hr.max()**2) / (valing_results['mse'] /
valing_results['batch_sizes']))
valing_results['ssim'] = valing_results[
'ssims'] / valing_results['batch_sizes']
val_bar.set_description(
desc=
'[Stats on validation set] PSNR: %.4f dB SSIM: %.4f' %
(valing_results['psnr'], valing_results['ssim']))
writer.add_scalars('stats/val', valing_results, epoch)
# Saving validation results
save_val_stats(cfg, epoch, valing_results)
netG.eval()
out_path = '../output/training_results/SRF_' + str(
default_config['upscale_factor']) + '/'
if not os.path.exists(out_path):
os.makedirs(out_path)
val_bar = tqdm(val_loader)
valing_results = {'mse': 0, 'ssims': 0, 'psnr': 0, 'ssim': 0}
val_images = []
for val_step, (lr, hr) in enumerate(val_bar):
lr = lr.to(device1)
hr = hr.to(device2)
with torch.no_grad():
sr = netG(lr).to(device2)
valing_results['mse'] += ((sr - hr)**2).data.mean()
valing_results['ssims'] += ssim(sr, hr).data.item()
if val_step < 20:
val_images.extend([
display_transform()(lr[0].cpu()),
display_transform()(hr.data.cpu().squeeze(0)),
display_transform()(sr.data.cpu().squeeze(0))
])
netG.train()
valing_results[
'mse'] = valing_results['mse'] / val_loader.__len__()
valing_results['psnr'] = 10 * log10(1 / valing_results['mse'])
valing_results[
'ssim'] = valing_results['ssims'] / val_loader.__len__()