torch_enhance.losses

class torch_enhance.losses.VGG(conv_index: str = '22')

VGG/Perceptual Loss

Parameters:conv_index (str) – Convolutional layer in VGG model to use as perceptual output

forward(sr: torch.Tensor, hr: torch.Tensor) → torch.Tensor

Compute VGG/Perceptual loss between Super-Resolved and High-Resolution

Parameters:

• sr (torch.Tensor) – Super-Resolved model output tensor
• hr (torch.Tensor) – High-Resolution image tensor

Returns:

loss – Perceptual VGG loss between sr and hr

Return type:

torch.Tensor