Research map / Point clouds / Point cloud classification
Exploring Point-BEV Fusion for 3D Point Cloud Object Tracking With Transformer
Zhipeng Luo, Changqing Zhou, Liang Pan, Gongjie Zhang, Tianrui Liu, Yueru Luo, Haiyu Zhao, Ziwei Liu, Shijian Lu
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 Relation-Aware Sampling instead of random sampling (ablation) | PTTR · KITTI SOT (average) | Success / Precision: 52.9/69.0 -> 58.4/77.8 | Improved |
| add RAS for pedestrian class (ablation) | PTTR · KITTI Pedestrian | Success / Precision: 36.6/59.9 -> 50.9/81.6 | Improved |
| add Point Relation Transformer instead of cosine similarity (ablation) | PTTR · KITTI SOT (average) | Success: 50.5 -> 58.4 | Improved |
| add Prediction Refinement Module (ablation) | PTTR · KITTI SOT (average) | Success: 56.7 -> 58.4 | Improved |
| add PRT and PRM together to base (ablation) | PTTR · KITTI SOT (average) | Success: 29.8 -> 58.4 | Improved |
| add offset attention to Relation Attention (ablation) | PTTR · KITTI SOT (average) | Success: 45.6 -> 52.2 | Improved |
| add L2 feature normalization to Relation Attention (ablation) | PTTR · KITTI SOT (average) | Success: 45.6 -> 57.0 | Improved |
| add BEV branch fusion to point branch (ablation) | PTTR++ · KITTI SOT (average) | Success / Precision: 56.7/75.2 -> 63.8/81.0 | Improved |
| add point branch fusion to BEV branch (ablation) | PTTR++ · KITTI SOT (average) | Success / Precision: 58.6/77.4 -> 63.8/81.0 | Improved |
| global attention instead of point-wise attention for fusion on point branch (ablation) | PTTR++ · KITTI SOT (average) | Success: 63.8 -> 61.9 | Hurt |
| global attention instead of addition for fusion on BEV branch (ablation) | PTTR++ · KITTI SOT (average) | Success: 62.7 -> 63.1 | Mixed |
| match BEV features at 1x instead of 4x downsampling (ablation) | PTTR++ · KITTI SOT (average) | Success / Precision: 59.3/76.3 -> 63.8/81.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MLVSNet — MLVSNet: Multi-Level Voting Siamese Network for 3D Visual Tracking (ICCV 2021) · 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
- PTTR: Relational 3D Point Cloud Object Tracking With Transformer (CVPR 2022 · similarity 0.94)
- PTT: Point-Trajectory Transformer for Efficient Temporal 3D Object Detection (CVPR 2024 · similarity 0.88)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.88)
- Towards Category Unification of 3D Single Object Tracking on Point Clouds (ICLR 2024 · similarity 0.88)
- Modeling Continuous Motion for 3D Point Cloud Object Tracking (AAAI 2024 · similarity 0.88)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.88)
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.
