Research map / 3D vision / Human pose estimation
Learning Realistic Human Reposing Using Cyclic Self-Supervision With 3D Shape, Pose, and Appearance Consistency
Soubhik Sanyal, Alex Vorobiov, Timo Bolkart, Matthew Loper, Betty Mohler, Larry S. Davis, Javier Romero, Michael J. Black
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 |
|---|---|---|---|
| add shape loss (ablation) | SPICE · DeepFashion | FID: 8.7 -> 7.8 | Improved |
| add pose loss (ablation) | SPICE · DeepFashion | FID: 8.4 -> 7.8 | Improved |
| add appearance loss (ablation) | SPICE · DeepFashion | FID: 9.9 -> 7.8 | Improved |
| condition discriminator on colored mesh renderings (ablation) | SPICE · DeepFashion | FID: 10.0 -> 7.8 | Improved |
| add shape loss (ablation) | SPICE · DeepFashion | LPIPS: 0.166 -> 0.164 | No clear effect |
| add pose loss (ablation) | SPICE · DeepFashion | LPIPS: 0.165 -> 0.164 | No clear effect |
| add appearance loss (ablation) | SPICE · DeepFashion | LPIPS: 0.164 -> 0.164 | No clear effect |
| condition discriminator on colored mesh renderings (ablation) | SPICE · DeepFashion | LPIPS: 0.167 -> 0.164 | Improved |
| use self-supervised SPICE instead of unpaired baseline PGSPT | unpaired reposing · DeepFashion | FID: 29.9 -> 7.8 | Improved |
| compute FID with raw reference vs JPEG-80 generated images (mismatched compression) | FID evaluation · DeepFashion | FID: 6.9 -> 12.1 | Hurt |
Closest papers by meaning
- Self-supervised 3D Representation Learning of Dressed Humans from Social Media Videos (TPAMI 2022 · similarity 0.89)
- Self-Supervised 3D Human Pose Estimation via Part Guided Novel Image Synthesis (CVPR 2020 · similarity 0.86)
- SelfPose3d: Self-Supervised Multi-Person Multi-View 3d Pose Estimation (CVPR 2024 · similarity 0.86)
- LoopReg: Self-supervised Learning of Implicit Surface Correspondences, Pose and Shape for 3D Human Mesh Registration (NeurIPS 2020 · similarity 0.86)
- MultiPly: Reconstruction of Multiple People from Monocular Video in the Wild (CVPR 2024 · similarity 0.85)
- MagicPose: Realistic Human Poses and Facial Expressions Retargeting with Identity-aware Diffusion (ICML 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.
