Research map / Point clouds / Point cloud classification
DPDist: Comparing Point Clouds Using Deep Point Cloud Distance
Dahlia Urbach, Yizhak Ben-Shabat, Michael Lindenbaum
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 |
|---|---|---|---|
| increase global Gaussian grid from 4^3 to 8^3 (ablation) | DPDist (local 3^3) · ModelNet40 chair reconstruction | Chamfer L1: 0.130189 -> 0.071805 | Improved |
| increase global Gaussian grid from 8^3 to 16^3 (ablation) | DPDist (local 3^3) · ModelNet40 chair reconstruction | Chamfer L1: 0.071805 -> 0.058943 | Improved |
| increase global Gaussian grid from 8^3 to 16^3 (ablation) | DPDist (local 3^3) · ModelNet40 chair reconstruction | Normal consistency: 0.794560 -> 0.809462 | Improved |
| increase local patch grid from 1^3 to 3^3 (ablation) | DPDist (global 8^3) · ModelNet40 chair reconstruction | Chamfer L1: 0.103590 -> 0.071805 | Improved |
| increase local patch grid from 3^3 to 5^3 (ablation) | DPDist (global 8^3) · ModelNet40 chair reconstruction | Normal consistency: 0.794560 -> 0.739863 | Hurt |
| increase local patch grid from 3^3 to 5^3 (ablation) | DPDist (global 8^3) · ModelNet40 chair reconstruction | Chamfer L1: 0.071805 -> 0.072714 | No clear effect |
| use DPDist instead of Chamfer distance for translation detection | DPDist · Sydney Urban cars | transformation distance (mean): 0.02879 -> 0.01385 | Improved |
| use DPDist instead of EMD for translation detection | DPDist · Sydney Urban cars | transformation distance (mean): 0.02381 -> 0.01385 | Improved |
| use DPDist instead of partial Hausdorff PH8 | DPDist · Sydney Urban cars | transformation distance (mean): 0.00863 -> 0.01385 | Hurt |
| use DPDist instead of Chamfer as PCRNet training loss | PCRNet · ModelNet40 chair registration, N=64,128,1024 | success ratio | Improved |
| use DPDist loss instead of Chamfer when the two clouds are identical samples | PCRNet · ModelNet40 chair registration, N=128 identical clouds | success ratio | No clear effect |
| use DPDist instead of Chamfer as auto-encoder loss | PointNet auto-encoder · ModelNet40, N=128 | point coverage | Hurt |
Closest papers by meaning
- Point-Set Distances for Learning Representations of 3D Point Clouds (ICCV 2021 · similarity 0.88)
- DeepMapping: Unsupervised Map Estimation From Multiple Point Clouds (CVPR 2019 · similarity 0.87)
- InfoCD: A Contrastive Chamfer Distance Loss for Point Cloud Completion (NeurIPS 2023 · similarity 0.87)
- Deep Learning for 3D Point Clouds: A Survey (TPAMI 2020 · similarity 0.87)
- APML: Adaptive Probabilistic Matching Loss for Robust 3D Point Cloud Reconstruction (NeurIPS 2025 · similarity 0.86)
- A Linear Time and Space Local Point Cloud Geometry Encoder via Vectorized Kernel Mixture (VecKM) (ICML 2024 · similarity 0.85)
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.
