AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Unsupervised Domain Adaption With Pixel-Level Discriminator for Image-Aware Layout Generation

Chenchen Xu, Min Zhou, Tiezheng Ge, Yuning Jiang, Weiwei Xu

CVPR 2023 · Image generation › Autoregressive and latent generation · approach: Layout generation · 15 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
Replace CGL-GAN discriminator with pixel-level discriminator on shallow features and drop Gaussian blurCGL-GAN (PDA-GAN) · CGL-DatasetR_com (lower better): 35.77 -> 33.55Improved
Replace CGL-GAN discriminator with pixel-level discriminator on shallow features and drop Gaussian blurCGL-GAN (PDA-GAN) · CGL-DatasetR_shm and R_sub (lower better): R_shm 15.47 -> 12.77; R_sub 0.805 -> 0.688Improved
Replace CGL-GAN discriminator with pixel-level discriminator and drop Gaussian blurCGL-GAN (PDA-GAN) · CGL-DatasetR_ove (lower better): 0.0233 -> 0.0290Hurt
Replace CGL-GAN discriminator with pixel-level discriminator and drop Gaussian blurCGL-GAN (PDA-GAN) · CGL-DatasetR_ali (lower better): 0.0098 -> 0.0105Hurt
Pixel-level discriminator instead of global discriminator (DA) (ablation)PDA-GAN · CGL-DatasetR_com, R_shm, R_sub, R_ove: best DA R_com 34.07 vs PDA 33.55; R_ove 0.0350 vs 0.0290Improved
Global discriminator with large loss weight (ablation)DA global discriminator · CGL-Datasetlayout generation success: DA-6.0 and DA-1.0 fail (R_ove ~9); DA-0.01 R_ove 4.7764Hurt
PatchGAN-style patch-level discriminator instead of pixel-level output (ablation)PDA-GAN discriminator · CGL-DatasetR_com, R_shm, R_sub: e.g. 88*60 patch: R_com 32.65 (better) but R_shm 13.35 and R_sub 0.735 vs pixel 12.77 and 0.688Hurt
Attach PD to deep (fourth residual block) features instead of shallow first-block features (ablation)PDA-GAN · CGL-DatasetR_com, R_shm, R_sub: R_com 34.22 vs 33.55; R_shm 13.97 vs 12.77; R_sub 0.770 vs 0.688Hurt
Attach PD to fused multi-level features instead of shallow first-block features (ablation)PDA-GAN · CGL-DatasetR_com, R_shm, R_sub: R_com 35.36 vs 33.55; R_sub 0.817 vs 0.688Hurt
Add one-target label smoothing (0 to 0.2 on non-inpainted pixels) (ablation)PDA-GAN · CGL-DatasetR_shm, R_sub, R_ove, R_und: R_shm 14.04 -> 12.77; R_sub 0.718 -> 0.688; R_ove 0.0346 -> 0.0290Improved
One-source label smoothing (only 1 to 0.8) instead of one-target (ablation)PDA-GAN · CGL-DatasetR_com and R_ali: R_com 32.20 vs 33.55 (better); R_ali 0.0085 vs 0.0105 (better) but R_shm 15.23 vs 12.77 and R_ove 0.0431 vs 0.0290 worseMixed

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