Research map / Image generation / Text-to-image diffusion
StyleT2I: Toward Compositional and High-Fidelity Text-to-Image Synthesis
Zhiheng Li, Martin Renqiang Min, Kai Li, Chenliang Xu
Check 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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use StyleT2I instead of the strongest reported baseline | Text-to-image synthesis · CelebA-HQ unseen compositions | R-Precision: 0.484 -> 0.625 | Improved |
| Use StyleT2I instead of the strongest reported baseline | Text-to-image synthesis · CUB unseen test | R-Precision: 0.145 -> 0.264 | Improved |
| Pretrain the generator on the larger cross-dataset corpus | StyleT2I · CelebA-HQ unseen compositions | R-Precision: 0.625 -> 0.698 | Improved |
| Pretrain the generator on the larger cross-dataset corpus | StyleT2I · CUB unseen test | FID: 20.53 -> 19.19 | Hurt |
| Add CLIP-guided Contrastive Loss (ablation) | StyleT2I · CUB | R-Precision: 0.205 -> 0.264 | Improved |
| Add norm penalty (ablation) | StyleT2I · CUB | FID: 23.86 -> 19.19 | Improved |
| Add norm penalty (ablation) | StyleT2I · CUB | R-Precision: 0.333 -> 0.264 | Hurt |
| Add Spatial Constraint (ablation) | StyleT2I · CUB | R-Precision: 0.246 -> 0.264 | Improved |
| Add Compositional Attribute Adjustment (ablation) | StyleT2I · CUB | R-Precision: 0.238 -> 0.264 | Improved |
| Fine-tune CLIP on the dataset training split (ablation) | StyleT2I · CUB | R-Precision: 0.145 -> 0.264 | Improved |
| Use norm threshold 16 instead of 8 (ablation) | StyleT2I · CelebA-HQ | R-Precision: 0.625 -> 0.815 | Improved |
| Use norm threshold 16 instead of 8 (ablation) | StyleT2I · CelebA-HQ | FID: 17.46 -> 21.35 | Hurt |
Later papers that compared with this method
1 later accepted paper reports StyleT2I as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Attribute-Centric Compositional Text-to-Image Generation (IJCV 2025 · better on 6 of 6 shared results)
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.
- DAE-GAN — DAE-GAN: Dynamic Aspect-Aware GAN for Text-to-Image Synthesis (ICCV 2021) · better on 3 of 3 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
- Enhancing Vision-Language Compositional Understanding with Multimodal Synthetic Data (CVPR 2025 · similarity 0.88)
- Progressive Compositionality in Text-to-Image Generative Models (ICLR 2025 · similarity 0.87)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.87)
- Discriminative Probing and Tuning for Text-to-Image Generation (CVPR 2024 · similarity 0.87)
- Progressive Text-to-Image Diffusion with Soft Latent Direction (AAAI 2024 · similarity 0.87)
- Noise Diffusion for Enhancing Semantic Faithfulness in Text-to-Image Synthesis (CVPR 2025 · similarity 0.87)
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.
