Research map / 3D vision / Human pose estimation
Learning Positive-Incentive Point Sampling in Neural Implicit Fields for Object Pose Estimation
Yifei Shi, Boyan Wan, Xin Xu, K. Xu
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 PIPS sampling instead of near-surface sampling (ablation) | full method · ShapeNet-C | 5deg5cm accuracy: 0.51 -> 0.62 | Improved |
| add PIPS-C certainty filtering of sample points (ablation) | full method · ShapeNet-C | 5deg5cm accuracy: 0.53 -> 0.62 | Improved |
| add PIPS-S geometric stability sampling (ablation) | full method · ShapeNet-C | 5deg5cm accuracy: 0.57 -> 0.62 | Improved |
| add anisotropic variance in pseudo ground truth instead of scalar variance (ablation) | full method · ShapeNet-C | 5deg5cm accuracy: 0.55 -> 0.62 | Improved |
| use SO(3)-equivariant convolutional implicit network instead of 3D-GCN (ablation) | full method · ShapeNet-C | 5deg5cm accuracy: 0.49 -> 0.62 | Improved |
| use SO(3)-equivariant network instead of 3D-GCN on held-out poses (ablation) | full method · ShapeNet-C holdout pose | 5deg5cm accuracy: 0.38 -> 0.61 | Improved |
| add RGB features (ablation) | full method · ShapeNet-C | 5deg5cm accuracy: 0.59 -> 0.62 | Improved |
| increase PIPS-C voxel size from 8 to 32 (ablation) | PIPS-C · ShapeNet-C | 5deg5cm accuracy: 0.64 -> 0.58 | Hurt |
| decrease PIPS-C voxel size from 8 to 4 (ablation) | PIPS-C · ShapeNet-C | 5deg5cm accuracy: 0.64 -> 0.49 | Hurt |
| lower target sparsity from 0.1 to 0.05 (ablation) | PIPS-S · ShapeNet-C | 5deg5cm accuracy: 0.62 -> 0.52 | Hurt |
| raise positive-sample threshold from 0.5 to 0.9 (ablation) | PIPS · ShapeNet-C | 5deg5cm accuracy: 0.62 -> 0.49 | Hurt |
| use PIPS method instead of AG-Pose (ablation) | AG-Pose · NOCS-REAL275 | 5deg2cm accuracy: 0.57 -> 0.63 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- SecondPose — SecondPose: SE(3)-Consistent Dual-Stream Feature Fusion for Category-Level Pose Estimation (CVPR 2024) · better on 4 of 5 shared results
- GDRNPP — GDRNPP: A Geometry-Guided and Fully Learning-Based Object Pose Estimator (TPAMI 2025) · better on 1 of 1 shared result
- ZebraPose — ZebraPose: Coarse To Fine Surface Encoding for 6DoF Object Pose Estimation (CVPR 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Implicit-PDF: Non-Parametric Representation of Probability Distributions on the Rotation Manifold (ICML 2021 · similarity 0.87)
- LEAP: Liberate Sparse-View 3D Modeling from Camera Poses (ICLR 2024 · similarity 0.87)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.87)
- Implicit Zoo: A Large-Scale Dataset of Neural Implicit Functions for 2D Images and 3D Scenes (NeurIPS 2024 · similarity 0.86)
- CRISP: Object Pose and Shape Estimation with Test-Time Adaptation (CVPR 2025 · similarity 0.86)
- IPoD: Implicit Field Learning with Point Diffusion for Generalizable 3D Object Reconstruction from Single RGB-D Images (CVPR 2024 · 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.
