Research map / Image generation / Text-to-image diffusion
Wasserstein Style Distribution Analysis and Transform for Stylized Image Generation
Xi Yu, Xiang Gu, Zhihao Shi, Jian Sun
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 changed | Where | Measured | Effect |
|---|---|---|---|
| raise threshold tau from 0.0 to 0.2 (transform style-relevant channels) (ablation) | WSDT · StyleDrop and InstaStyle evaluation sets | Style Score: 0.476 -> 0.546 | Improved |
| raise threshold tau from 0.0 to 0.2 (ablation) | WSDT · StyleDrop and InstaStyle evaluation sets | Content Score: 0.298 -> 0.285 | Hurt |
| raise tau from 0.2 to 0.4 (ablation) | WSDT · StyleDrop and InstaStyle evaluation sets | Style Score: 0.546 -> 0.546 | No clear effect |
| raise tau from 0.2 to 0.4 (ablation) | WSDT · StyleDrop and InstaStyle evaluation sets | Content Score: 0.285 -> 0.283 | Hurt |
| transform all channels instead of adaptive selection (tau 0.2) (ablation) | WSDT · StyleDrop and InstaStyle evaluation sets | Content Score: 0.285 -> 0.281 | Hurt |
| transform all channels instead of adaptive selection (tau 0.2) (ablation) | WSDT · StyleDrop and InstaStyle evaluation sets | Style Score: 0.546 -> 0.547 | No clear effect |
| use WSDT instead of the strongest baseline StyleAligned | SDXL · StyleDrop and InstaStyle evaluation sets | GM Metric: 10.53 -> 5.52 | Improved |
| use WSDT instead of VSP | SDXL · StyleDrop and InstaStyle evaluation sets | User Preference: 0.246 -> 0.494 | Improved |
| use WSDT instead of StyleAligned | SDXL · StyleDrop and InstaStyle evaluation sets | Style Score: 0.538 -> 0.546 | Improved |
| use WSDT instead of VSP | SDXL · StyleDrop and InstaStyle evaluation sets | Content Score: 0.282 -> 0.285 | Improved |
| apply AdaIN on noisy latents instead of WSDT feature transform (ablation) | SDXL · qualitative | style alignment | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CSGO — CSGO: Content-Style Composition in Text-to-Image Generation (NeurIPS 2025) · better on 4 of 4 shared results
- DEADiff — DEADiff: An Efficient Stylization Diffusion Model with Disentangled Representations (CVPR 2024) · better on 4 of 4 shared results
- InstaStyle — InstaStyle: Inversion Noise of a Stylized Image is Secretly a Style Adviser (ECCV 2024) · better on 4 of 4 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
- Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · similarity 0.89)
- Z*: Zero-shot Style Transfer via Attention Reweighting (CVPR 2024 · similarity 0.89)
- Conditional Balance: Improving Multi-Conditioning Trade-Offs in Image Generation (CVPR 2025 · similarity 0.88)
- Attention Distillation: A Unified Approach to Visual Characteristics Transfer (CVPR 2025 · similarity 0.88)
- Free-Lunch Color-Texture Disentanglement for Stylized Image Generation (NeurIPS 2025 · similarity 0.88)
- SigStyle: Signature Style Transfer via Personalized Text-to-Image Models (AAAI 2025 · similarity 0.88)
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.
