What does "Fréchet Inception Distance" (FID) measure in GANs?

Fréchet Inception Distance (FID) is a metric used to evaluate the quality of generated images by measuring how similar they are to real images in terms of their statistical feature representations. Specifically, FID calculates the distance between the feature vectors of real and generated samples, which are typically extracted from a pre-trained Inception model. The lower the FID score, the closer the generated images are to the real images in the feature space, indicating better image quality and diversity. This makes FID a widely adopted evaluation tool in the field of Generative Adversarial Networks (GANs), helping researchers assess the effectiveness of their models based on generated image similarity to real images.