Research map / Image generation / Autoregressive and latent generation
Efficient Semantic Image Synthesis via Class-Adaptive Normalization
Zhentao Tan, Dongdong Chen, Qi Chu, Menglei Chai, Jing Liao, Mingming He, Lu Yuan, Gang Hua, Nenghai Yu
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 class-adaptive normalization CLADE | SPADE generator · ADE20k | Params (M): 96.5 -> 71.4 | Improved |
| replace SPADE with CLADE | SPADE generator · ADE20k | FLOPs (G): 181.3 -> 42.2 | Improved |
| replace SPADE with CLADE | SPADE generator · ADE20k | mIoU: 36.28 -> 35.43 | Mixed |
| replace SPADE with CLADE | SPADE generator · Cityscapes | mIoU: 61.95 -> 60.44 | Hurt |
| replace SPADE with CLADE | SPADE generator · ADE20k CPU runtime | runtime (s): 4.104s -> 1.710s | Improved |
| add intra-class positional encoding (CLADE-ICPE) to CLADE (ablation) | CLADE · Cityscapes | FID: 50.62 -> 42.39 | Improved |
| add intra-class positional encoding (CLADE-ICPE) to CLADE (ablation) | CLADE · ADE20k | FID: 30.48 -> 28.69 | Improved |
| add positional encoding by concatenation with input mask (+disti) (ablation) | CLADE · ADE20k | FID: 30.48 -> 31.75 | Hurt |
| add positional encoding by concatenation with normalized features (+distf) (ablation) | CLADE · ADE20k | FID: 30.48 -> 31.13 | Hurt |
| use SPADE on low-res layers and CLADE on high-res layers (ablation) | CLADE · Cityscapes | mIoU: 60.44 -> 62.14 | Improved |
| use CLADE instead of lightweight SPADE with similar FLOPs | SPADE-light · ADE20k | mIoU: 26.29 -> 35.43 | Improved |
| replace SPADE with CLADE in SEAN | SEAN · CelebAMask-HQ | FID: 24.30 -> 20.35 | Improved |
Closest papers by meaning
- Semantic Image Synthesis With Spatially-Adaptive Normalization (CVPR 2019 · similarity 0.92)
- Retrieval-Based Spatially Adaptive Normalization for Semantic Image Synthesis (CVPR 2022 · similarity 0.91)
- Semantic Probability Distribution Modeling for Diverse Semantic Image Synthesis (TPAMI 2022 · similarity 0.89)
- PLACE: Adaptive Layout-Semantic Fusion for Semantic Image Synthesis (CVPR 2024 · similarity 0.89)
- SSMG: Spatial-Semantic Map Guided Diffusion Model for Free-Form Layout-to-Image Generation (AAAI 2024 · similarity 0.87)
- Stochastic Conditional Diffusion Models for Robust Semantic Image Synthesis (ICML 2024 · similarity 0.86)
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.
