Research map / Image generation / Editing and style transfer
Progressive Pose Attention Transfer for Person Image Generation
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 |
|---|---|---|---|
| Use progressive pose-attentional transfer blocks | Person image generation · Market-1501 | SSIM: 0.291 -> 0.311 | Helped |
| Use progressive pose-attentional transfer blocks | Person image generation · DeepFashion | PCKh: 0.94 -> 0.96 | Helped |
| Replace residual blocks with PATBs (ablation) | Generator · Market-1501 | PCKh: 0.84 -> 0.91 | Helped |
| Use 5 PATBs rather than 9 (ablation) | PATN generator · Market-1501 | PCKh: 0.94 -> 0.91 | Hurt |
| Increase PATB count from 9 to 13 (ablation) | PATN generator · Market-1501 | PCKh: 0.94 -> 0.93 | Hurt |
| Remove the addition operation from each PATB (ablation) | PATN generator · Market-1501 | PCKh: 0.94 -> 0.92 | Hurt |
| Remove the concatenation operation from each PATB (ablation) | PATN generator · Market-1501 | PCKh: 0.94 -> 0.92 | Hurt |
| Remove discriminator residual blocks (ablation) | PATN · Market-1501 | PCKh: 0.94 -> 0.93 | Hurt |
| Remove the addition operation from each PATB (ablation) | PATN generator · DeepFashion | PCKh: 0.96 -> 0.93 | Hurt |
| Remove the concatenation operation from each PATB (ablation) | PATN generator · DeepFashion | PCKh: 0.96 -> 0.95 | Hurt |
| Remove discriminator residual blocks (ablation) | PATN · DeepFashion | PCKh: 0.96 -> 0.94 | Hurt |
| Augment reduced Market-1501 training data with PATN-generated images | ResNet-50 re-ID model · Market-1501, 0.3 real-data portion | re-ID accuracy: 41.5 -> 56.5 | Helped |
Closest papers by meaning
- Pose Guided Person Image Generation (NeurIPS 2017 · similarity 0.91)
- Unsupervised Person Image Synthesis in Arbitrary Poses (CVPR 2018 · similarity 0.89)
- MUST-GAN: Multi-Level Statistics Transfer for Self-Driven Person Image Generation (CVPR 2021 · similarity 0.88)
- Deformable GANs for Pose-Based Human Image Generation (CVPR 2018 · similarity 0.87)
- Disentangled Person Image Generation (CVPR 2018 · similarity 0.86)
- IMAGPose: A Unified Conditional Framework for Pose-Guided Person Generation (NeurIPS 2024 · 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.