AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Responsible Disclosure of Generative Models Using Scalable Fingerprinting

Ning Yu, Vladislav Skripniuk, Dingfan Chen, Larry S. Davis, Mario Fritz

ICLR 2022 · Image generation › Autoregressive and latent generation · approach: Generalizable fake image detection · 20 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use filter-modulated fingerprints rather than passing fingerprints through the generator input (ablation)StyleGAN2 fingerprinting · LSUN Catbit accuracy: 0.500 -> 0.996Improved
Modulate filters separately with latent and fingerprint embeddings (ablation)StyleGAN2 fingerprinting · LSUN Catbit accuracy: 0.996 -> 0.869Hurt
Modulate filters using concatenated latent and fingerprint embeddings (ablation)StyleGAN2 fingerprinting · LSUN BedroomFID: 20.50 -> 64.91Hurt
Modulate all generator layers instead of a single 16×16 layer (ablation)StyleGAN2 fingerprinting · CelebAbit accuracy: 0.993 -> 0.991No clear effect
Modulate all generator layers instead of a single 8×8 layer (ablation)StyleGAN2 fingerprinting · LSUN Bedroombit accuracy: 0.950 -> 0.993Improved
Add the consistency loss (ablation)Fingerprint GAN with adversarial and fingerprint reconstruction losses · CelebAFID: 12.84 -> 12.24Improved
Add latent-code reconstruction loss (ablation)Fingerprint GAN with adversarial, fingerprint, and consistency losses · CelebAFID: 12.24 -> 11.50Improved
Use only adversarial and fingerprint reconstruction losses, omitting consistency and latent reconstruction (ablation)Fingerprint GAN · CelebAbit accuracy: 0.999 vs 0.991No clear effect
Use on-the-fly fingerprint sampling instead of a finite training fingerprint set (ablation)StyleGAN2 fingerprinting · CelebAtesting bit accuracy: 0.752 -> 0.991Improved
Train with a finite set of 10 fingerprints rather than sampling fingerprints on the fly (ablation)StyleGAN2 fingerprinting · CelebAtesting bit accuracy: 0.991 -> 0.512Hurt
Train with perturbed-image augmentation for immunization (ablation)Fingerprint GAN · CelebA under cropping, Gaussian noise, and combined perturbationsbit accuracyImproved
Use proactive fingerprint attribution instead of passive classifiersDeep fake detection and attribution · CelebA open world, 10 GAN sourcesaccuracy: 0.102 -> 1.000Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy