Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Learning Analytical Posterior Probability for Human Mesh 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 |
|---|---|---|---|
| Use the analytical posterior over joint rotations rather than the prior-only approach | Human mesh recovery on 3DPW · 3DPW | MPJPE: 74.1 -> 68.3 | Helped |
| Use the analytical posterior over joint rotations | Human mesh recovery on Human3.6M · Human3.6M | MPJPE: 47.1 -> 45.7 | Helped |
| Use the posterior framework for multi-view and IMU fusion | TotalCapture · TotalCapture, mv + IMUs | MPJPE: 24.6 -> 23.5 | Helped |
| Add the 3D keypoints branch (ablation) | Ours on Human3.6M · Human3.6M | MPJPE: 76.5 -> 45.7 | Helped |
| Add prior distribution parameters F to the model (ablation) | Ours on Human3.6M · Human3.6M | MPJPE: 63.6 -> 45.7 | Helped |
| Retain prior F at test time rather than setting it to zero (ablation) | Ours on Human3.6M · Human3.6M | MPJPE: 58.7 -> 45.7 | Helped |
| Use an intermediate feature rather than a late feature for F and β (ablation) | Ours on Human3.6M · Human3.6M | MPJPE: 48.6 -> 45.7 | Helped |
| Use an intermediate feature rather than an early feature for F and β (ablation) | Ours on Human3.6M · Human3.6M | MPJPE: 46.5 -> 45.7 | Helped |
| Fuse the learned prior with keypoint-derived bone directions instead of using keypoints alone (ablation) | Ours on Human3.6M · Human3.6M | MPJPE: 63.6 -> 45.7 | Helped |
| Use the posterior strategy rather than the no-prior keypoint strategy as keypoint noise increases (ablation) | Human3.6M noise test · Human3.6M | MPJPE: 61.3 -> 48.5; 62.7 -> 49.6; 65.0 -> 51.4; 71.5 -> 54.5; 80.4 -> 59.0 | Helped |
| Use AGORA NIKI results as a comparison for the proposed method | Human mesh recovery on AGORA · AGORA test set | MPJPE: NIKI 67.3; Ours 74.4 | Mixed |
Closest papers by meaning
- Reconstructing Humans with a Biomechanically Accurate Skeleton (CVPR 2025 · similarity 0.87)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 2024 · similarity 0.86)
- Incorporating Test-Time Optimization into Training with Dual Networks for Human Mesh Recovery (NeurIPS 2024 · similarity 0.86)
- Nonisotropic Gaussian Diffusion for Realistic 3D Human Motion Prediction (CVPR 2025 · similarity 0.85)
- KITRO: Refining Human Mesh by 2D Clues and Kinematic-tree Rotation (CVPR 2024 · similarity 0.84)
- ADHMR: Aligning Diffusion-based Human Mesh Recovery via Direct Preference Optimization (ICML 2025 · similarity 0.84)
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.