AI Paper Atlas

Research map / Image generation / GANs

Discriminator Feature-Based Inference by Recycling the Discriminator of GANs

Duhyeon Bang, Seoungyoon Kang, Hyunjung Shim

IJCV 2020 · Image generation › GANs · approach: Large-scale GAN latent analysis · 12 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 DFI (discriminator features + latent reconstruction) instead of VAEinference mapping · CIFAR10FID: 234.03 -> 31.54Improved
use DFI (WGAN-GP) instead of ALICEinference mapping · CelebAFID: 16.36 -> 8.73Improved
use DFI (WGAN-GP) instead of ALI/BiGANinference mapping · Fashion MNISTFID: 69.99 -> 11.85Improved
use DFI with SNGAN base instead of other GANsDFI · CIFAR10FID: 31.54 (WGAN-GP) vs 61.88 (SNGAN)Hurt
use LSGAN base instead of DCGAN for DFIDFI · CelebAFID: 12.86 -> 24.99Hurt
add latent reconstruction loss instead of image reconstruction loss (ablation)DFI · CelebAFID / LPIPSMixed
add latent optimization (50 iterations) after inference mapping (ablation)DFI · CelebAuser study / color fidelityImproved
use discriminator features instead of pre-trained VGG16 Pool5 features (ablation)DFI · CelebALPIPS, FID, user studyImproved
add GAP of intermediate layer Actv64-1 to discriminator features (ablation)DFI (SNGAN) · CelebAFID: 8.91 -> 8.54Improved
add GAP of Actv128-0 to the base discriminator feature (ablation)DFI (SNGAN) · CelebAFID: 8.91 -> 10.34Hurt
combine GAP features from all three intermediate layers (ablation)DFI (SNGAN) · CelebALPIPS / FID: LPIPS 0.1931 -> 0.1887; FID 8.91 -> 10.32Mixed
use 4x4 average pooling instead of GAP (1x1) on Actv64-1 (ablation)DFI (SNGAN) · CelebALPIPS / FID: LPIPS 0.1903 -> 0.1835; FID 8.54 -> 8.26Improved

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