Research map / 3D vision / Human pose estimation
DPCN++: Differentiable Phase Correlation Network for Versatile Pose Registration
Zexi Chen, Yiyi Liao, Haozhe Du, Haodong Zhang, Xuecheng Xu, Haojian Lu, Rong Xiong, Yue Wang
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 |
|---|---|---|---|
| use learned feature extractors with phase correlation instead of learning-free PC | DPCN++ · 2D simulation (heterogeneous) | Acc rotation 1deg (%): 72.3 -> 100 | Improved |
| replace direct pose regression (DAM) with differentiable phase correlation solver | DPCN++ · AG scene (a) l2sat | Acc x 10px (%): 55.4 -> 96.9 | Improved |
| add explicit intermediate representation (DAM vs RPR) | regression baseline · 2D simulation (heterogeneous) | Acc rotation 1deg (%): 78.3 -> 80.8 | Improved |
| increase bandwidth from 64 to 128 (ablation) | DPCN++ · 3DMatch Home 1 | registration success rate (%): 93.5 -> 96.2 | Improved |
| increase bandwidth from 64 to 128 (ablation) | DPCN++ · 3DMatch Study | registration success rate (%): 92.3 -> 94.7 | Improved |
| increase bandwidth from 64 to 128 (ablation) | DPCN++ · 3DMatch | time (ms): 17 -> 37 | Hurt |
| add translation invariance step to the solver (ablation) | DPCN++ · MVP, unseen translations | registration error | Improved |
| add certification to TEASER | TEASER · 3DMatch Home 1 | registration success rate (%): 92.9 -> 94.1 | Mixed |
| train on heterogeneous data without prior conversion of representations (DPCN++ vs DPCN++*) | DPCN++ · point cloud/mesh/SDF | pose error | No clear effect |
| use partial instead of complete point clouds | TEASER · MVP | pose error | Hurt |
Closest papers by meaning
- Recurrent Multi-View Alignment Network for Unsupervised Surface Registration (CVPR 2021 · similarity 0.86)
- Deep Global Registration (CVPR 2020 · similarity 0.85)
- Disentangle-then-Align: Non-Iterative Hybrid Multimodal Image Registration via Cross-Scale Feature Disentanglement (CVPR 2026 · similarity 0.85)
- Locally Aware Piecewise Transformation Fields for 3D Human Mesh Registration (CVPR 2021 · similarity 0.85)
- CARL: A Framework for Equivariant Image Registration (CVPR 2025 · similarity 0.85)
- MRC-Net: 6-DoF Pose Estimation with MultiScale Residual Correlation (CVPR 2024 · 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.
