Research map / Point clouds / Segmentation and detection
PARTNER: Level up the Polar Representation for LiDAR 3D Object Detection
Ming Nie, Yujing Xue, Chunwei Wang, Chaoqiang Ye, Hang Xu, Xinge Zhu, Qingqiu Huang, Michael Bi Mi, Xinchao Wang, Li Zhang
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 global representation re-alignment (GRR) to polar detector (ablation) | polar CenterPoint · Waymo val vehicle | L1 mAP (%): 70.31 -> 72.62 | Improved |
| add geometry-aware adaptive (GA) module on top of GRR (ablation) | PARTNER · Waymo val vehicle | L1 mAP (%): 72.62 -> 75.35 | Improved |
| apply GRR to Cartesian detector (ablation) | Cartesian CenterPoint · Waymo val vehicle | L1 mAP (%): 73.51 -> 73.04 | Hurt |
| add auxiliary segmentation and regression losses only (ablation) | GA module · Waymo val vehicle | L1 mAP (%): 72.62 -> 72.48 | No clear effect |
| add geometric clues to attention (ablation) | GA module · Waymo val vehicle | L1 mAP (%): 74.34 -> 75.35 | Improved |
| use axial attention for re-alignment instead of GRR (ablation) | polar detector · Waymo val vehicle | L1 mAP (%): 73.33 -> 73.41 | No clear effect |
| use Swin or channel-wise attention for re-alignment instead of GRR (ablation) | polar detector · Waymo val vehicle | L1 mAP (%): 73.33 -> 73.62 | No clear effect |
| regress in polar coordinates instead of Cartesian (ablation) | PARTNER · Waymo val vehicle | L1 mAP (%): 75.35 -> 75.02 | Hurt |
| use SID or LID instead of uniform range partition (ablation) | PARTNER · Waymo val vehicle | L1 mAP (%): 75.35 -> 74.32 | Hurt |
| use polar instead of Cartesian voxels at coarse resolution | CenterPoint · Waymo val vehicle 5x voxel size | L2 mAPH (%): 35.6 -> 46.0 | Improved |
| use polar instead of Cartesian voxels at fine resolution | CenterPoint · Waymo val vehicle 1x voxel size | L2 mAPH (%): 65.5 -> 62.1 | Hurt |
| apply PARTNER on PointPillars for streaming detection | PolarStream · Waymo val streaming, 8 sectors | L2 mAP (%): 58.6 -> 63.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PolarStream — PolarStream: Streaming Object Detection and Segmentation with Polar Pillars (NeurIPS 2021) · better on 2 of 2 shared results
- PV-RCNN — PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020) · 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
- PARTNER: Level Up the Polar Representation for 3D Object Detection (IJCV 2026 · similarity 0.97)
- PolarStream: Streaming Object Detection and Segmentation with Polar Pillars (NeurIPS 2021 · similarity 0.92)
- PolarFormer: Multi-Camera 3D Object Detection with Polar Transformer (AAAI 2023 · similarity 0.88)
- GBlobs: Explicit Local Structure via Gaussian Blobs for Improved Cross-Domain LiDAR-based 3D Object Detection (CVPR 2025 · similarity 0.87)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.87)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 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.
