AI Paper Atlas

Research map / Image generation / Editing and style transfer

Contrastive Learning for Diverse Disentangled Foreground Generation

Yuheng Li, Yijun Li, Jingwan Lu, Eli Shechtman, Yong Jae Lee, Krishna Kumar Singh

ECCV 2022 · Image generation › Editing and style transfer · approach: Diffusion face generation and editing · 10 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 a second latent code set (known and unknown) instead of single code (ablation)ContrasFill · Face foreground generationKFFA: 74.71 -> 83.66Improved
add a second latent code set (known and unknown) instead of single code (ablation)ContrasFill · Car foreground generationKFFA: 63.09 -> 82.96Improved
use bi-modulated convolution instead of concatenating codes (ablation)ContrasFill · Face foreground generationKFFA: 48.99 -> 83.06Improved
use bi-modulation instead of sequential u-k modulated convolution sets (ablation)ContrasFill · Face foreground generationKFFA: 78.19 -> 83.06Hurt
use contrastive loss instead of reprediction loss (ablation)ContrasFill · Face foreground generationFID: 14.41 -> 8.41Improved
use contrastive loss instead of reprediction loss (ablation)ContrasFill · Face foreground generationLPIPS: 0.120 -> 0.075Hurt
ContrasFill vs CoModGAN on known-factor diversityContrasFill · Face foreground generationKFFA: 52.41 -> 83.66Improved
ContrasFill vs BAT-Fill image qualityContrasFill · Face foreground generationFID: 15.08 -> 8.36Improved
sample known code via disentangled space vs moving along discovered directionContrasFill · Face KFFA over 1,000 contextsKFFA: 56.00 -> 82.03Improved
fix known code and vary unknown codesContrasFill · Face KFFA over 1,000 contextsKFFA: 82.03 -> 26.15Hurt
object-shaped mask: ContrasFill vs CoModGANContrasFill · Face foreground generation (object mask)KFFA: 51.19 -> 83.39Improved
object-shaped mask: ContrasFill vs CoModGANContrasFill · Bird foreground generation (object mask)LPIPS: 0.030 -> 0.101Improved

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