AI Paper Atlas

Research map / Image generation / GANs

How good is my GAN?

Konstantin Shmelkov, Cordelia Schmid, Karteek Alahari

ECCV 2018 · Image generation › GANs · approach: GAN latent optimization and evaluation · 409 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 SNGAN instead of WGAN-GP (10M)GAN evaluation · CIFAR100 generated imagesGAN-train accuracy: 26.7 -> 45.0Improved
use SNGAN instead of DCGANGAN evaluation · CIFAR10 generated imagesGAN-test accuracy: 58.2 -> 87.3Improved
increase WGAN-GP size from 2.5M to 10M parameters (ablation)WGAN-GP · CIFAR10 generated imagesGAN-train accuracy: 76.1 -> 79.5Improved
increase WGAN-GP size from 2.5M to 10M parameters (ablation)WGAN-GP · CIFAR100 generated imagesGAN-test accuracy: 4.3 -> 40.4Improved
add salt and pepper noise (1% to 20% of pixels) to CIFAR10 (ablation)GAN-test measure · CIFAR10 corrupted training setGAN-test accuracy: 82% -> 15%Hurt
subsample CIFAR10 train set to a random 5k subset with Gaussian noise (ablation)GAN-train measure · CIFAR10 corrupted training setGAN-train accuracy: 91% -> 80%Hurt
train SNGAN at 128px instead of 64pxSNGAN · ImageNet generated imagesGAN-train top-1 accuracy: 3 -> 9.3Improved
replace CNN classifier with random forests (ablation)GAN-train/GAN-test · CIFAR100 generated imagesGAN ranking: 15.2%, 19.5% for SNGANNo clear effect
add 50k SNGAN images to 2.5k real imagesCIFAR10 classifier · CIFAR10 augmentation, reduced real setclassification accuracy: 73.4 -> 67.0Hurt
add 50k SNGAN images to 10k real imagesCIFAR100 classifier · CIFAR100 augmentation, reduced real setclassification accuracy: 51.5 -> 45.5Hurt
add 50k real images to 50k GAN imagesCIFAR100 classifier · CIFAR100 augmentationclassification accuracyNo clear effect
add 2.5k or 5k real images to 50k GAN imagesCIFAR10 and CIFAR100 classifier · augmentation with full-data SNGANclassification accuracyImproved

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