def metrics_evaluation(self):
# 3D metrics including mse, cs and psnr
g_fake_unNorm = self.ct_unGaussian(self.G_fake)
g_real_unNorm = self.ct_unGaussian(self.G_real)
self.metrics_Mse = Metrics.Mean_Squared_Error(g_fake_unNorm, g_real_unNorm)
self.metrics_CosineSimilarity = Metrics.Cosine_Similarity(g_fake_unNorm, g_real_unNorm)
self.metrics_PSNR = Metrics.Peak_Signal_to_Noise_Rate(g_fake_unNorm, g_real_unNorm, PIXEL_MAX=1.0)
def forward(self):
# output is [B D H W]
self.G_fake = self.netG(self.G_input)
# visual object should be [B 1 D H W]
self.G_fake_D = torch.unsqueeze(self.G_fake, 1)
if not self.training:
if self.opt.CT_MEAN_STD[0] == 0:
self.G_fake = torch.clamp(self.G_fake, 0, 1)
elif self.opt.CT_MEAN_STD[0] == 0.5:
self.G_fake = torch.clamp(self.G_fake, -1, 1)
else:
raise NotImplementedError()
# input of Discriminator is [B 1 D H W]
self.G_real_D = torch.unsqueeze(self.G_real, 1)
if self.conditional_D:
self.G_condition_D = self.G_input.unsqueeze(1).expand_as(self.G_real_D)
# map
self.G_Map_real_F = self.transition(self.output_map(self.ct_unGaussian(self.G_real_D), 2).squeeze(1))
self.G_Map_fake_F = self.transition(self.output_map(self.ct_unGaussian(self.G_fake_D), 2).squeeze(1))
self.G_Map_real_S = self.transition(self.output_map(self.ct_unGaussian(self.G_real_D), 4).squeeze(1))
self.G_Map_fake_S = self.transition(self.output_map(self.ct_unGaussian(self.G_fake_D), 4).squeeze(1))
if self.training:
for i in self.multi_view:
out_map = self.output_map(self.ct_unGaussian(self.G_real_D), i + 1).squeeze(1)
out_map = self.ct_Gaussian(self.transition(out_map))
setattr(self, 'G_Map_{}_real'.format(i), out_map)
out_map = self.output_map(self.ct_unGaussian(self.G_fake_D), i + 1).squeeze(1)
out_map = self.ct_Gaussian(self.transition(out_map))
setattr(self, 'G_Map_{}_fake'.format(i), out_map)
# metrics
g_fake_unNorm = self.ct_unGaussian(self.G_fake)
g_real_unNorm = self.ct_unGaussian(self.G_real)
self.metrics_Mse = Metrics.Mean_Squared_Error(g_fake_unNorm, g_real_unNorm)
self.metrics_CosineSimilarity = Metrics.Cosine_Similarity(g_fake_unNorm, g_real_unNorm)
self.metrics_PSNR = Metrics.Peak_Signal_to_Noise_Rate(g_fake_unNorm, g_real_unNorm, PIXEL_MAX=1.0)
def forward(self):
# output is [B 1 D H W]
self.G_fake_D = self.netG(self.noise)
# visual object should be [B D H W]
self.G_fake = torch.squeeze(self.G_fake_D, 1)
# input of Discriminator is [B 1 D H W]
self.G_real_D = torch.unsqueeze(self.G_real, 1)
# value should clip to 0-1 when inference
if not self.training:
self.G_fake = torch.clamp(self.G_fake, 0, 1)
# map
self.G_map_real = self.transition(self.output_map(self.G_real, 1))
if self.G_fake_D.dim() == 4:
self.G_map_fake = self.transition(self.output_map(
self.G_fake_D, 1))
elif self.G_fake_D.dim() == 5:
self.G_map_fake = self.output_map(self.G_fake_D, 2)
self.G_map_fake = self.transition(self.G_map_fake.squeeze_(1))
else:
raise ValueError
# metrics
self.metrics_Mse = Metrics.Mean_Squared_Error(self.G_fake, self.G_real)
self.metrics_CosineSimilarity = Metrics.Cosine_Similarity(
self.G_fake, self.G_real)