Research map / 3D vision / Human pose estimation
Aligning Silhouette Topology for Self-Adaptive 3D Human Pose Recovery
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 |
|---|---|---|---|
| adapt with pixel-level L2 silhouette loss (B1) instead of no adaptation (ablation) | Ours(S) · 3DPW S->R | PA-MPJPE: 134.7 -> 129.6 | Helped |
| replace L2 with chamfer spatial alignment loss on raw silhouettes (B2) (ablation) | B1 · 3DPW S->R | PA-MPJPE: 129.6 -> 106.2 | Helped |
| add L2 loss on topological skeleton to B2 (B3) (ablation) | B2 · 3DPW S->R | PA-MPJPE: 106.2 -> 105.5 | No clear effect |
| use chamfer-inspired topological alignment loss instead of L2 on skeleton (ablation) | B3 · 3DPW S->R | PA-MPJPE: 105.5 -> 95.1 | Helped |
| self-supervised silhouette adaptation vs pre-adaptation | Ours · 3DPW S->R | PA-MPJPE: 134.7 -> 95.1 | Helped |
| silhouette adaptation instead of OpenPose 2D pseudo labels (A1) | SPIN · UAP-H3M eps 16/255 | MPJPE: 151.2 -> 125.2 | Helped |
| add silhouette fitting loss to 2D pose adaptation (A2 vs A1) | SPIN · UAP-H3M eps 16/255 | MPJPE: 151.2 -> 136.5 | Helped |
| adapt with 2D pose pseudo labels (A1) on weak perturbation | SPIN · UAP-H3M eps 4/255 | MPJPE: 65.8 -> 64.5 | No clear effect |
| our silhouette adaptation at weak perturbation | Ours(R) · UAP-H3M eps 4/255 | MPJPE: 67.7 -> 63.6 | Helped |
| our silhouette adaptation at strongest shift | Ours(R) · UAP-H3M eps 16/255 | PA-MPJPE: 91.2 -> 79.4 | Helped |
| silhouette adaptation vs A2 on LR domain 32 | SPIN · LR-3DPW 32 | MPJPE: 147.5 -> 134.2 | Helped |
| silhouette adaptation vs A1 on LR domain 96 | SPIN · LR-3DPW 96 | MPJPE: 100.2 -> 99.8 | No clear effect |
Closest papers by meaning
- Self-Supervised 3D Human Pose Estimation via Part Guided Novel Image Synthesis (CVPR 2020 · similarity 0.87)
- ManiPose: Manifold-Constrained Multi-Hypothesis 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.87)
- Learning Basis Representation to Refine 3D Human Pose Estimations (AAAI 2019 · similarity 0.87)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.87)
- TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation (CVPR 2024 · similarity 0.87)
- FactorizedHMR: A Hybrid Framework for Video Human Mesh Recovery (NeurIPS 2026 · 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.