AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Attribute-Centric Compositional Text-to-Image Generation

Yuren Cong, Martin Renqiang Min, Li Erran Li, Bodo Rosenhahn, Michael Ying Yang

IJCV 2025 · Image generation › Text-to-image diffusion · approach: Personalized and region-controlled generation · 25 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 independent CLIP-based matching discriminator D_m (ablation)ACTIG · CUBFID: 18.3 -> 13.9Improved
add independent CLIP-based matching discriminator D_m (ablation)ACTIG · CUBR-Precision: 15.2 -> 21.8Improved
add attribute-centric contrastive loss (ablation)ACTIG · CUBFID: 13.9 -> 12.8Improved
add attribute-centric contrastive loss (ablation)ACTIG · CUBR-Precision: 21.8 -> 25.6Improved
add image-free training with attribute-centric feature augmentation (ablation)ACTIG · CUBR-Precision: 25.6 -> 28.7Improved
add image-free training (ablation)ACTIG · CUBFID: 12.8 -> 12.5No clear effect
finetune CLIP with norm penalty instead of pre-trained CLIP (ablation)ACTIG · CelebA-HQR-Precision: 57.1 -> 65.0Improved
finetune CLIP with norm penalty instead of pre-trained CLIP (ablation)ACTIG · CUBR-Precision: 19.5 -> 28.7Improved
add norm penalty to CLIP finetuning (ablation)ACTIG · CelebA-HQR-Precision: 61.2 -> 65.0Improved
add norm penalty to CLIP finetuning (ablation)ACTIG · CUBFID: 12.3 -> 12.5No clear effect
increase attribute count in input text from <=2 to 3ACTIG · CelebA-HQ (K=100)R-Precision: 41.3 -> 67.0Improved
increase attribute count from 5 to >=6ACTIG · CelebA-HQ (K=100)R-Precision: 67.5 -> 63.8Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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