Conceptperceptual loss
Perceptual Loss
A loss function based on differences in intermediate-layer features of a pretrained feature-extraction network. It optimizes the texture and semantic similarity of reconstructed images rather than per-pixel error, and is commonly used in super-resolution and image restoration.
Timeline (last 90 days)
Oct 21
The perceptual loss was proposed by Johnson et al.; it feeds the reconstructed image and the real image into a pretrained feature extraction network and compares differences in intermediate-layer feature maps.
Unverified50%