AI Paper Atlas

Research map / Image generation / GANs

Towards Discovery and Attribution of Open-World GAN Generated Images

Sharath Girish, Saksham Suri, Sai Saketh Rambhatla, Abhinav Shrivastava

ICCV 2021 · Image generation › GANs · approach: GAN latent optimization and evaluation · 62 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
add OOD detection, merge, refine and iterations to the single-iteration pipeline (k=209) (ablation)GAN discovery pipeline · GAN discovery (20 sources, k=209)Avg. Purity / NMI: 0.795 -> 0.861 purity, 0.609 -> 0.724 NMIImproved
add the merge step after clustering (ablation)GAN discovery pipeline · GAN discovery, iteration 1 stagesnumber of clusters, Avg. Purity: 512 -> 391 clusters, 0.793 -> 0.792 purity, 0.682 -> 0.689 NMIImproved
add the SVM refine step, evaluated on discovered samples (ablation)GAN discovery pipeline · GAN discovery, iteration 1 stagesAvg. Purity / NMI: 0.792 -> 0.849 purity, 0.689 -> 0.838 NMIImproved
add the refine step, evaluated on all samples (ablation)GAN discovery pipeline · GAN discovery, iteration 1 stagesAvg. Purity: 0.792 -> 0.772 purity, 0.689 -> 0.720 NMIHurt
run more iterations (1 to 4) (ablation)GAN discovery pipeline · GAN discovery by iterationAvg. Purity / sources discovered: 0.772 -> 0.861 purity, 16/20 -> 20/20 sourcesImproved
add one GAN per dataset to the labeled set (ablation)GAN discovery pipeline · GAN discovery (labeled set variants)Avg. Purity: 0.497 -> 0.897Improved
add a second GAN per dataset to the labeled set (ablation)GAN discovery pipeline · GAN discovery (labeled set variants)Avg. Purity: 0.897 -> 0.954Improved
use WTA hash distance instead of cosine distance for OOD (ablation)OOD detection step · GAN discovery, 4 iterationsnet OOD accuracy: 89.33 -> 90.65 at iteration 1, 88.27 -> 90.33 at iteration 4Improved
add JPEG and blur augmentations to the method (ablation)GAN discovery pipeline · GAN discovery (20 sources)Avg. Purity / NMI: 0.8216 -> 0.7696 purity, 0.6552 -> 0.6249 NMIHurt
use cluster majority voting for real/fake detection instead of per-sample predictionreal/fake classifier · Real/fake detection, original setupaccuracy: 92.56% -> 98.62%Improved
use cluster majority voting for real/fake detection in new real plus new GANs setupreal/fake classifier · Real/fake detection, new real + new GANsaccuracy: 89.09% -> 83.33%Hurt

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