Research map / Point clouds / Point cloud classification
RMS-FlowNet++: Efficient and Robust Multi-scale Scene Flow Estimation for Large-Scale Point Clouds
Ramy Battrawy, René Schuster, Didier Stricker
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 |
|---|---|---|---|
| reduce correspondence set Kp from 33 to 16 (ablation) | RMS-FlowNet · FT3Ds / KITTIs | Acc3DR: 95.60 -> 95.15 (FT3Ds); 95.80 -> 91.72 (KITTIs) | Hurt |
| add graph representation to flow embedding (ablation) | RMS-FlowNet++ (FPS) · FT3Ds | Acc3DR: 95.15 -> 96.85 | Improved |
| omit the decoder in feature extraction (ablation) | RMS-FlowNet++ (FPS) · FT3Ds | Acc3DR: 96.85 -> 97.63 | Improved |
| add 2nd embedding step in feature space (ablation) | RMS-FlowNet++ (FPS) · KITTIs | Acc3DR: 94.57 -> 95.52 | Improved |
| add bidirectional similarity map (ablation) | RMS-FlowNet++ (FPS) · FT3Ds | Acc3DR: 97.77 -> 98.09 | Improved |
| add bidirectional similarity map (ablation) | RMS-FlowNet++ (FPS) · KITTIs | Acc3DR: 95.52 -> 97.63 | Improved |
| add spatial encoding to stacked attention (ablation) | RMS-FlowNet++ (FPS) · FT3Ds | Acc3DR: 95.13 -> 96.84 | Improved |
| add reference-frame feature f_i^t to attention input (ablation) | RMS-FlowNet++ (FPS) · FT3Ds | Acc3DR: 96.84 -> 97.20 | Improved |
| add residual connection to stacked attention (ablation) | RMS-FlowNet++ (FPS) · FT3Ds | Acc3DR: 97.20 -> 98.09 | Improved |
| increase Kp from 8 to 16 (FPS) (ablation) | RMS-FlowNet++ · FT3Ds | Acc3DR: 97.70 -> 98.09 | Improved |
| increase Kp from 16 to 20 (FPS) (ablation) | RMS-FlowNet++ · FT3Ds | Acc3DR: 98.09 -> 98.07 | No clear effect |
| train with random sampling (Kp=20) instead of FPS (Kp=16) (ablation) | RMS-FlowNet++ · KITTIs | Acc3DR: 97.63 -> 97.67 | 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.
- PV-RAFT — PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Clouds (CVPR 2021) · better on 1 of 1 shared result
- PointPWC-Net — PointPWC-Net: Cost Volume on Point Clouds for (Self-)Supervised Scene Flow Estimation (ECCV 2020) · better on 1 of 1 shared result
- FlowNet3D — FlowNet3D: Learning Scene Flow in 3D Point Clouds (CVPR 2019) · 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
- FlowNet3D: Learning Scene Flow in 3D Point Clouds (CVPR 2019 · similarity 0.90)
- IHNet: Iterative Hierarchical Network Guided by High-Resolution Estimated Information for Scene Flow Estimation (ICCV 2023 · similarity 0.89)
- PointPWC-Net: Cost Volume on Point Clouds for (Self-)Supervised Scene Flow Estimation (ECCV 2020 · similarity 0.89)
- HPLFlowNet: Hierarchical Permutohedral Lattice FlowNet for Scene Flow Estimation on Large-Scale Point Clouds (CVPR 2019 · similarity 0.89)
- DELFlow: Dense Efficient Learning of Scene Flow for Large-Scale Point Clouds (ICCV 2023 · similarity 0.88)
- FlowMamba: Learning Point Cloud Scene Flow with Global Motion Propagation (AAAI 2025 · similarity 0.87)
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.
