Research map / Image generation / Editing and style transfer
SEAN: Image Synthesis With Semantic Region-Adaptive Normalization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 normalization with SEAN's per-region style-conditioned normalization | Semantic image synthesis generator · CelebAMask-HQ | FID: 22.43 -> 17.66 | Helped |
| Replace SPADE normalization with SEAN's per-region style-conditioned normalization | Semantic image synthesis generator · CityScapes | FID: 60.51 -> 50.38 | Helped |
| Replace SPADE normalization with SEAN's per-region style-conditioned normalization | Semantic image synthesis generator · ADE20K | FID: 34.65 -> 24.84 | Helped |
| Replace Pix2PixHD style conditioning with SEAN | Semantic image synthesis generator · Façades | FID: 22.34 -> 19.82 | Helped |
| Use a unified encoder instead of separate per-layer style encoders (ablation) | SEAN · CelebAMask-HQ | FID: 18.17 -> 17.66; 17.98 -> 17.66 | Helped |
| Use a unified encoder instead of separate per-layer style encoders (ablation) | SEAN · CelebAMask-HQ | PSNR: 19.70 -> 18.74; 18.86 -> 18.74 | Hurt |
| Add downsampling to the style encoder (ablation) | SEAN · CelebAMask-HQ | FID: 18.67 -> 17.66 | Helped |
| Add downsampling to the style encoder (ablation) | SEAN · CelebAMask-HQ | PSNR: 18.32 -> 18.74 | Helped |
| Add learnable noise inputs (ablation) | SEAN · CelebAMask-HQ | FID: 18.05 -> 17.66 | Helped |
| Add learnable noise inputs (ablation) | SEAN · CelebAMask-HQ | PSNR: 18.49 -> 18.74 | Helped |
| Remove the mask-conditioned branch from SEAN normalization (ablation) | SEAN · Image synthesis | visual quality | Hurt |
Closest papers by meaning
- Editing in Style: Uncovering the Local Semantics of GANs (CVPR 2020 · similarity 0.90)
- Semantic Image Synthesis With Spatially-Adaptive Normalization (CVPR 2019 · similarity 0.89)
- Learning Semantic-aware Normalization for Generative Adversarial Networks (NeurIPS 2020 · similarity 0.89)
- EditGAN: High-Precision Semantic Image Editing (NeurIPS 2021 · similarity 0.88)
- SemanticStyleGAN: Learning Compositional Generative Priors for Controllable Image Synthesis and Editing (CVPR 2022 · similarity 0.88)
- High-Resolution Image Synthesis and Semantic Manipulation With Conditional GANs (CVPR 2018 · 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.