Research map / Point clouds / Registration and matching
Locally Aware Piecewise Transformation Fields for 3D Human Mesh Registration
Check 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 PTF-FC pose initialization before fitting (ablation) | SMPL/SMPL+D registration · CAPE validation | registration error (mm): 27.5 mm -> 23.1 mm; 27.7 mm -> 23.1 mm | Helped |
| Use PTF-FC instead of IPNet | clothed-human registration · CAPE validation | outer registration error (mm): 28.2 mm -> 23.1 mm | Helped |
| Use PTF-FC instead of IPNet | under-cloth body registration · CAPE validation | inner registration error (mm): 28.3 mm -> 23.1 mm | Helped |
| Estimate pose with point-aligned local features and correspondence fitting rather than direct global-feature rotation regression | SMPL pose estimation · CAPE validation | per-vertex error (mm): 74.4 mm -> 34.1 mm | Helped |
| Use ConvONet rather than IFNet as encoder (ablation) | IPNet decoder reconstruction · CAPE validation | mIoU: 86.7% -> 87.8% | Helped |
| Add random rotation augmentation to ConvONet inputs (ablation) | IPNet decoder reconstruction · CAPE validation | mIoU: 87.8% -> 88.6% | Helped |
| Replace IPNet decoder with PTF-FC | human surface reconstruction · CAPE validation | mIoU: 88.6% -> 89.7% | Helped |
| Add direct correspondence prediction to IPNet while retaining posed-space occupancy inference (ablation) | IPNet · CAPE validation | outer CD: 0.0151 -> 0.0153 | Hurt |
| Replace PTF with a 4-layer MLP (ablation) | PTF-FC · CAPE validation | mIoU: 89.7% -> 88.1% | Hurt |
| Use PTF-Piecewise instead of IPNet | human surface reconstruction · CAPE validation | mIoU: 88.6% -> 89.4% | Helped |
Closest papers by meaning
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.90)
- LoopReg: Self-supervised Learning of Implicit Surface Correspondences, Pose and Shape for 3D Human Mesh Registration (NeurIPS 2020 · similarity 0.89)
- Rectified Point Flow: Generic Point Cloud Pose Estimation (NeurIPS 2025 · similarity 0.88)
- Local-consistent Transformation Learning for Rotation-invariant Point Cloud Analysis (CVPR 2024 · similarity 0.87)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 2024 · similarity 0.87)
- Reconstructing Humans with a Biomechanically Accurate Skeleton (CVPR 2025 · 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.