Research map / 3D vision / Human pose estimation
The Devil Is in the Pose: Ambiguity-Free 3D Rotation-Invariant Learning via Pose-Aware Convolution
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 |
|---|---|---|---|
| Add pose-aware convolution with APPF and factorized dynamic kernels | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.1 -> 92.3 | Helped |
| Add pose-aware convolution | PaRI-Conv · ShapeNetPart z/SO(3) | mIoU (%): 82.2 -> 83.8 | Helped |
| Use PCA-based local primal axis rather than global-center axis | PaRI-Conv · ScanObjectNN OBJ_BG | accuracy (%): 78.1 -> 83.3 | Helped |
| Use APPF rather than rotation matrix and translation vector (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.2 -> 92.3 | Helped |
| Add full APPF direction and pose information rather than PPF (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.5 -> 92.3 | Helped |
| Add direction information to APPF (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.4 -> 92.3 | Helped |
| Use factorized dynamic kernels rather than APPF-feature concatenation (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.3 -> 92.3 | Helped |
| Use factorized dynamic kernels rather than RSCNN kernels (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.2 -> 92.3 | Helped |
| Use factorized dynamic kernels rather than PointConv kernels (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.2 -> 92.3 | Helped |
| Use factorized dynamic kernels rather than PAConv kernels (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 89.6 -> 92.3 | Helped |
| Add final EdgeConv after pose-aware kernel aggregation (ablation) | PaRI-Conv · ModelNet40 SO(3)/SO(3) | accuracy (%): 91.7 -> 92.3 | Helped |
| Use a normal and local-barycenter LRF rather than a normal and global-center LRF (ablation) | PaRI-Conv · ModelNet40 | accuracy (%): 91.7 -> 92.3 | Helped |
Closest papers by meaning
- PaRot: Patch-Wise Rotation-Invariant Network via Feature Disentanglement and Pose Restoration (AAAI 2023 · similarity 0.92)
- Enhancing Rotation-Invariant 3D Learning with Global Pose Awareness and Attention Mechanisms (AAAI 2026 · similarity 0.90)
- Sparse Steerable Convolutions: An Efficient Learning of SE(3)-Equivariant Features for Estimation and Tracking of Object Poses in 3D Space (NeurIPS 2021 · similarity 0.88)
- 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction (NeurIPS 2024 · similarity 0.86)
- LEAP: Liberate Sparse-View 3D Modeling from Camera Poses (ICLR 2024 · similarity 0.86)
- Learning with 3D rotations, a hitchhiker's guide to SO(3) (ICML 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.