Research map / 3D vision / Human pose estimation
Dense Pose Transfer
Natalia Neverova, Riza Alp Guler, Iasonas Kokkinos
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 |
|---|---|---|---|
| condition on DensePose instead of body keypoints (ablation) | predictive module · DeepFashion | SSIM: 0.762 -> 0.792 | Improved |
| condition on DensePose instead of foreground mask (ablation) | predictive module · DeepFashion | SSIM: 0.747 -> 0.792 | Improved |
| condition on DensePose instead of body part segmentation (ablation) | predictive module · DeepFashion | SSIM: 0.776 -> 0.792 | Improved |
| use DensePose as both input and target representation vs keypoints both (ablation) | predictive module · DeepFashion | MS-SSIM: 0.774 -> 0.821 | Improved |
| add blending (self-refinement) (ablation) | predictive module · DeepFashion | SSIM: 0.792 -> 0.793 | No clear effect |
| add warping branch to predictive + blending (ablation) | pipeline · DeepFashion | SSIM: 0.793 -> 0.789 | Hurt |
| add inpainting in UV space to warping (ablation) | pipeline · DeepFashion | SSIM: 0.789 -> 0.796 | Improved |
| add perceptual loss to L1 (ablation) | full pipeline · DeepFashion | SSIM: 0.796 -> 0.791 | Hurt |
| add style loss to L1 (ablation) | full pipeline · DeepFashion | Inception Score: 3.17 -> 3.67 | Improved |
| add style loss to L1 (ablation) | full pipeline · DeepFashion | SSIM: 0.796 -> 0.777 | Hurt |
| add GAN loss to L1 (ablation) | full pipeline · DeepFashion | SSIM: 0.796 -> 0.771 | Hurt |
| combine all four losses vs L1 only (ablation) | full pipeline · DeepFashion | Inception Score: 3.17 -> 3.61 | Improved |
Closest papers by meaning
- DensePose: Dense Human Pose Estimation in the Wild (CVPR 2018 · similarity 0.91)
- DenseRaC: Joint 3D Pose and Shape Estimation by Dense Render-and-Compare (ICCV 2019 · similarity 0.88)
- Dense Intrinsic Appearance Flow for Human Pose Transfer (CVPR 2019 · similarity 0.87)
- Synthesizing Images of Humans in Unseen Poses (CVPR 2018 · similarity 0.87)
- DenseReg: Fully Convolutional Dense Shape Regression In-The-Wild (CVPR 2017 · similarity 0.87)
- Can Generative Video Models Help Pose Estimation? (CVPR 2025 · similarity 0.87)
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.
