Research map / Image generation / GANs
Deformable GANs for Pose-Based Human 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 |
|---|---|---|---|
| Use deformable skip connections instead of standard U-Net skip connections (ablation) | Pose-conditioned person image generation · Market-1501 | DS: 0.595 -> 0.629 | Helped |
| Use deformable skip connections instead of standard U-Net skip connections (ablation) | Pose-conditioned person image generation · Market-1501 | SSIM: 0.256 -> 0.272 | Helped |
| Replace L1 loss with perceptual loss without nearest-neighbour search (ablation) | DSC generator · Market-1501 | DS: 0.629 -> 0.603 | No clear effect |
| Use nearest-neighbour loss instead of perceptual loss without nearest-neighbour search (ablation) | DSC generator · Market-1501 | DS: 0.603 -> 0.720 | Helped |
| Use nearest-neighbour loss instead of perceptual loss without nearest-neighbour search (ablation) | DSC generator · Market-1501 | SSIM: 0.276 -> 0.290 | Helped |
| Use the proposed full method rather than Ma et al. | Pose-conditioned person image generation · Market-1501 | DS: 0.39^* -> 0.72 | Helped |
| Use the proposed full method rather than Ma et al. | Pose-conditioned person image generation · Market-1501 | IS: 3.460 -> 3.185 | Hurt |
| Use the proposed full method rather than Ma et al. | Pose-conditioned person image generation · DeepFashion | IS: 3.090 -> 3.439 | Helped |
| Use the proposed full method rather than Ma et al. | Pose-conditioned person image generation · DeepFashion | SSIM: 0.762 -> 0.756 | Hurt |
| Augment Re-ID training data with generated pose variations | IDE + Euclidean · Market-1501 test set | mAP: 48.8 -> 55.9 | Helped |
Closest papers by meaning
- Unsupervised Person Image Synthesis in Arbitrary Poses (CVPR 2018 · similarity 0.90)
- Pose Guided Person Image Generation (NeurIPS 2017 · similarity 0.88)
- Constrained Generative Adversarial Networks for Interactive Image Generation (CVPR 2019 · similarity 0.88)
- Progressive Pose Attention Transfer for Person Image Generation (CVPR 2019 · similarity 0.87)
- Multistage Adversarial Losses for Pose-Based Human Image Synthesis (CVPR 2018 · similarity 0.87)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 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.