Research map / Point clouds / Point cloud classification
Dynamic 3D Scene Analysis by Point Cloud Accumulation
Shengyu Huang, Zan Gojcic, Jiahui Huang, Andreas Wieser, Konrad Schindler
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 |
|---|---|---|---|
| add foreground segmentation to exclude dynamic objects from registration (ablation) | sequential scene flow model · Waymo static EPE (cm) | EPE avg: 4.1 -> 2.9 | Improved |
| add foreground segmentation (ablation) | sequential scene flow model · Waymo dynamic EPE (cm) | EPE avg: 28.6 -> 19.7 | Improved |
| add motion segmentation to recover ego-motion of static foreground (ablation) | sequential scene flow model · Waymo static foreground EPE (cm) | EPE avg: 19.0 -> 2.1 | Improved |
| add motion segmentation to recover ego-motion of static foreground (ablation) | sequential scene flow model · nuScenes static foreground EPE (cm) | EPE avg: 19.9 -> 7.4 | Improved |
| add offset compensation for instance association (ablation) | sequential scene flow model · nuScenes instance association | WCov: +3.1 pp | Improved |
| add offset compensation for instance association (ablation) | sequential scene flow model · Waymo dynamic EPE (cm) | EPE avg: reduces EPE of dynamic parts by 3.7 cm | Improved |
| estimate ego-motion directly to target frame instead of chaining previous-frame poses (ablation) | sequential scene flow model · Waymo static part | EPE avg: 0.044 -> 0.029 | Improved |
| estimate ego-motion directly to target frame instead of chaining previous-frame poses (ablation) | sequential scene flow model · Waymo dynamic EPE | EPE avg: 0.171 -> 0.197 | Mixed |
| use spatio-temporal association with TubeNet instead of Kalman tracker (ablation) | sequential scene flow model · Waymo dynamic EPE | EPE avg: 0.586 -> 0.197 | Improved |
| use spatio-temporal association with TubeNet instead of Kalman tracker (ablation) | sequential scene flow model · nuScenes dynamic EPE | EPE avg: 1.238 -> 0.301 | Improved |
| use learned TubeNet instead of centroid translation (ablation) | motion modelling · nuScenes dynamic EPE | EPE avg: 0.553 -> 0.301 | Improved |
| add ICP refinement after the model | Ours · Waymo dynamic EPE | EPE avg: 0.197 -> 0.173 | Improved |
Closest papers by meaning
- Learning Scene Dynamics from Point Cloud Sequences (IJCV 2022 · similarity 0.89)
- Less Is More: Reducing Task and Model Complexity for 3D Point Cloud Semantic Segmentation (CVPR 2023 · similarity 0.88)
- LiDAR-Net: A Real-scanned 3D Point Cloud Dataset for Indoor Scenes (CVPR 2024 · similarity 0.88)
- CurveCloudNet: Processing Point Clouds with 1D Structure (CVPR 2024 · similarity 0.87)
- AS-Det: Active Sampling for Adaptive 3D Object Detection in Point Clouds (AAAI 2025 · similarity 0.87)
- Scaling Diffusion Models to Real-World 3D LiDAR Scene Completion (CVPR 2024 · 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.
