Research map / Image generation / Editing and style transfer
Pose Guided Person Image Generation
Read the paperCheck 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 coordinate embedding with concatenated pose heatmaps (ablation) | G1 · DeepFashion and Market-1501 | SSIM: 0.694 -> 0.735; 0.219 -> 0.304 | Helped |
| replace coordinate embedding with concatenated pose heatmaps (ablation) | G1 · DeepFashion and Market-1501 | IS: 2.395 -> 2.427; 2.568 -> 3.006 | Helped |
| use a separate heatmap encoder instead of concatenated pose heatmaps (ablation) | G1 · DeepFashion | SSIM: 0.735 -> 0.735 | No clear effect |
| use pose mask loss instead of standard L1 loss (ablation) | G1 · DeepFashion and Market-1501 | SSIM: 0.735 -> 0.779; 0.304 -> 0.340 | Helped |
| use pose mask loss instead of standard L1 loss (ablation) | G1 · DeepFashion and Market-1501 | IS: 2.427 -> 2.668; 3.006 -> 3.326 | Helped |
| add a second-stage refinement generator (ablation) | G1+D · DeepFashion | IS: 3.091 -> 3.090 | No clear effect |
| add a second-stage refinement generator (ablation) | G1+D · Market-1501 | SSIM: 0.283 -> 0.253 | Hurt |
| add a second-stage refinement generator (ablation) | G1+D · Market-1501 | mask-IS: 3.310 -> 3.435 | Helped |
| train the second-stage generator to produce a difference map rather than generate the target image from scratch | PG$^2$ · person pose transfer | training convergence | Helped |
| add a pose mask to weight foreground pixels in the loss (ablation) | G1 · Market-1501 | visual sharpness and color quality | Helped |
| add a second-stage refinement generator (ablation) | G1+D · DeepFashion | G2R user-study rate: 9.3% -> 14.9% | Helped |
| add a second-stage refinement generator (ablation) | G1+D · Market-1501 | G2R user-study rate: 11.1% -> 5.5% | Hurt |
Closest papers by meaning
- Progressive Pose Attention Transfer for Person Image Generation (CVPR 2019 · similarity 0.91)
- Unsupervised Person Image Synthesis in Arbitrary Poses (CVPR 2018 · similarity 0.91)
- Structure-aware Person Image Generation with Pose Decomposition and Semantic Correlation (AAAI 2021 · similarity 0.89)
- Person Image Synthesis via Denoising Diffusion Model (CVPR 2023 · similarity 0.88)
- IMAGPose: A Unified Conditional Framework for Pose-Guided Person Generation (NeurIPS 2024 · similarity 0.87)
- Coarse-to-Fine Latent Diffusion for Pose-Guided Person Image Synthesis (CVPR 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.