The Role of ImageNet Classes in Fréchet Inception Distance

Fréchet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID "looks at" i...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Kynkäänniemi, Tuomas, Karras, Tero, Aittala, Miika, Aila, Timo, Lehtinen, Jaakko
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 14.02.2023
Subjects
Affine transformations
Histograms
Seat belts
Online AccessGet full text

Cover

More Information
Summary:Fréchet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID "looks at" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-\(N\) classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation.
ISSN:2331-8422