Research map / Image generation / Editing and style transfer
Learning Semantic Person Image Generation by Region-Adaptive Normalization
Zhengyao Lv, Xiaoming Li, Xin Li, Fu Li, Tianwei Lin, Dongliang He, Wangmeng Zuo
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 per-region adaptive normalization with per-region style translation (ablation) | SPGNet · DeepFashion test | FID: 18.946 -> 12.613 | Improved |
| replace SEAN with SPADE (ablation) | SPGNet · DeepFashion test | FID: 12.613 -> 13.808 | Hurt |
| replace SEAN with CC-FPSE (ablation) | SPGNet · DeepFashion test | FID: 12.613 -> 13.686 | Hurt |
| add target semantic parsing maps to the pose encoder (ablation) | Baseline · DeepFashion test | FID: 18.946 -> 16.460 | Improved |
| add skeleton distance maps to the pose representation (ablation) | SPATN · DeepFashion semantic parsing | mIOU: 0.520 -> 0.539 | Improved |
| use SEAN directly without feature deformation (ablation) | person image generation · DeepFashion test | FID: 12.613 -> 15.993 | Hurt |
| jointly train the pre-trained SPATN and SPGNet (ablation) | two-stage SPGNet · DeepFashion test | FID: 12.243 -> 24.309 | Hurt |
| pre-train SPATN, freeze it, then train SPGNet (ablation) | two-stage SPGNet · DeepFashion test | FID: 12.243 -> 17.340 | Hurt |
| use region-adaptive style normalization rather than global normalization (ablation) | SPGNet · person image generation | LPIPS: The paper reports that Ours-Full has lower LPIPS than the compared variants. | Improved |
| predict target semantic maps and use them for per-region style translation | SPGNet · Market-1501 test | PCKh: 0.94 -> 0.97 | Improved |
Closest papers by meaning
- Unsupervised Person Image Generation With Semantic Parsing Transformation (CVPR 2019 · similarity 0.89)
- IMAGPose: A Unified Conditional Framework for Pose-Guided Person Generation (NeurIPS 2024 · similarity 0.88)
- Coarse-to-Fine Latent Diffusion for Pose-Guided Person Image Synthesis (CVPR 2024 · similarity 0.86)
- Advancing Pose-Guided Image Synthesis with Progressive Conditional Diffusion Models (ICLR 2024 · similarity 0.86)
- Semantic-Aware Human Object Interaction Image Generation (ICML 2024 · similarity 0.86)
- Disentangled Pose and Appearance Guidance for Multi-Pose Generation (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.
