kaira.losses.image.LPIPSLoss

Inheritance diagram for LPIPSLoss

class kaira.losses.image.LPIPSLoss

Bases: BaseLoss

Learned Perceptual Image Patch Similarity (LPIPS) Loss Module.

This module calculates the LPIPS loss between the input and the target. LPIPS uses deep features to measure perceptual similarity between images, which correlates better with human judgment than pixel-based metrics [Zhang et al., 2018].

Methods

__init__	Initialize the LPIPSLoss module.
forward	Forward pass through the LPIPSLoss module.

__init__()

Initialize the LPIPSLoss module.

forward(x: Tensor, target: Tensor) → Tensor

Forward pass through the LPIPSLoss module.

Parameters:

• x (torch.Tensor) – The input tensor.

• target (torch.Tensor) – The target tensor.

Returns:

The LPIPS loss between the input and the target.

Return type:

torch.Tensor