Research map / Point clouds / Point cloud classification
Self-Supervised Robust Scene Flow Estimation via the Alignment of Probability Density Functions
Read the paperCheck 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 |
|---|---|---|---|
| replace Chamfer Distance with Cauchy-Schwarz divergence (ablation) | Ours scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | EPE3D [m]: 0.170 -> 0.105 | Helped |
| replace Earth Mover’s Distance with Cauchy-Schwarz divergence (ablation) | Ours scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | EPE3D [m]: 0.192 -> 0.105 | Helped |
| replace Chamfer Distance with Cauchy-Schwarz divergence (ablation) | Ours scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | Acc3DS: 0.477 -> 0.633 | Helped |
| replace Earth Mover’s Distance with Cauchy-Schwarz divergence (ablation) | Ours scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | Acc3DS: 0.426 -> 0.633 | Helped |
| add graph Laplacian regularization with λ = 10 instead of no regularization (ablation) | CS scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | EPE3D[m]: 0.105 -> 0.096 | Helped |
| set graph Laplacian regularization weight to 100 rather than 10 (ablation) | CS scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | EPE3D[m]: 0.096 -> 0.109 | Hurt |
| set graph Laplacian regularization weight to 0.1 rather than 10 (ablation) | CS scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | EPE3D[m]: 0.096 -> 0.103 | Hurt |
| increase the number of input points (ablation) | CS scene-flow model · KITTI Scene Flow, trained on KITTIᵣ | EPE3D [m]: 0.369 (1,024) -> 0.184 (2,048) -> 0.171 (4,096) | Helped |
| use Silverman's rule of thumb or Improved Sheater-Jones bandwidth selection instead of the chosen bandwidth (ablation) | CS divergence · Kernel bandwidth ablation | scene-flow performance | Hurt |
| choose a very large or very small Gaussian kernel variance instead of a suitable bandwidth (ablation) | CS divergence · Kernel bandwidth ablation | scene-flow performance | Hurt |
| remove the refinement module from FLOT in self-supervised training (ablation) | FLOT · KITTI Scene Flow | scene-flow performance | Helped |
Closest papers by meaning
- Self-Point-Flow: Self-Supervised Scene Flow Estimation From Point Clouds With Optimal Transport and Random Walk (CVPR 2021 · similarity 0.93)
- 3DSFLabelling: Boosting 3D Scene Flow Estimation by Pseudo Auto-labelling (CVPR 2024 · similarity 0.87)
- ZeroFlow: Scalable Scene Flow via Distillation (ICLR 2024 · similarity 0.86)
- VoteFlow: Enforcing Local Rigidity in Self-Supervised Scene Flow (CVPR 2025 · similarity 0.86)
- FlowMamba: Learning Point Cloud Scene Flow with Global Motion Propagation (AAAI 2025 · similarity 0.86)
- ICP-Flow: LiDAR Scene Flow Estimation with ICP (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.