Frechet Inception Distance (FID)

Fréchet Inception Distance (FID) is a performance metric proposed by Heusel et al. in 2017 that calculates the distance between the feature vector of a real image and the feature vector of a fake image (generated by the generator). Lower FID scores indicate that the images generated by the generator are of higher quality and similar to the real image. FID is based on the feature vector of the image.