Research map / Image generation / Editing and style transfer
Context-Consistent Semantic Image Editing with Style-Preserved Modulation
Wuyang Luo, Su Yang, Hong Wang, Bo Long, Weishan Zhang
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 |
|---|---|---|---|
| replace SPADE with style-preserved modulation (SPM) (ablation) | SPMPGAN · ADE20K-Room (free-form) | FID: 23.27 -> 18.83 | Improved |
| replace SPADE with SPM (ablation) | SPMPGAN · ADE20K-Landscape (free-form) | FID: 34.71 -> 23.11 | Improved |
| replace SPADE with SPM (ablation) | SPMPGAN · Cityscapes (outpainting) | FID: 31.33 -> 27.63 | Improved |
| bypass normalization when generating context parameters (not use normalized maps) (ablation) | SPMPGAN · ADE20K-Landscape (free-form) | FID: 29.87 -> 23.11 | Improved |
| bypass normalization for context parameters (ablation) | SPMPGAN · ADE20K-Room (outpainting) | FID: 42.31 -> 41.52 | Improved |
| add progressive coarse-to-fine architecture (ablation) | SPMPGAN · ADE20K-Landscape (free-form) | FID: 25.87 -> 23.11 | Improved |
| add progressive architecture (ablation) | SPMPGAN · ADE20K-Room (extension) | FID: 33.87 -> 32.61 | Improved |
| increase SPADE network scale (SPADE-L) without SPM (ablation) | w SPADE · ADE20K-Room (free-form) | FID: 23.27 -> 24.11 | No clear effect |
| halve channels in SPM (SPMPGAN-S) (ablation) | SPMPGAN · ADE20K-Room (free-form) | FID: 18.83 -> 18.93 | No clear effect |
| SPMPGAN-S with fewer parameters vs SPADE-L (ablation) | SPMPGAN-S · ADE20K-Room (free-form) | FID: 24.11 -> 18.93 | Improved |
| SPMPGAN vs SESAME | SPMPGAN · ADE20K-Landscape (free-form) | FID: 30.30 -> 23.11 | Improved |
| SPMPGAN vs SESAME | SPMPGAN · Cityscapes (free-form) | LPIPS: 0.082 -> 0.084 | Hurt |
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.
- SESAME — SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects (ECCV 2020) · better on 12 of 12 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
- Diverse Semantic Image Synthesis via Probability Distribution Modeling (CVPR 2021 · similarity 0.87)
- PLACE: Adaptive Layout-Semantic Fusion for Semantic Image Synthesis (CVPR 2024 · similarity 0.87)
- Improving Editability in Image Generation with Layer-wise Memory (CVPR 2025 · similarity 0.86)
- Spin: Diffusion-based Semantic Image Painting Through Independent Information Injection (AAAI 2025 · similarity 0.86)
- SSMG: Spatial-Semantic Map Guided Diffusion Model for Free-Form Layout-to-Image Generation (AAAI 2024 · similarity 0.85)
- Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing (CVPR 2025 · similarity 0.85)
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.
