Research map / Point clouds / Segmentation and detection
PARTNER: Level Up the Polar Representation for 3D Object Detection
Ming Nie, Chunwei Wang, Hang Xu, 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 baseline · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 62.74 -> 65.02 | Improved |
| add geometry-aware adaptive module (GA) on top of GRR (ablation) | Polar + GRR · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 65.02 -> 67.84 | Improved |
| apply GRR to Cartesian detector (ablation) | Cartesian baseline · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 65.86 -> 65.38 | Hurt |
| add auxiliary supervision in GA (ablation) | Polar + GRR · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 65.02 -> 64.89 | No clear effect |
| add window attention in GA (ablation) | Polar + GRR · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 65.02 -> 65.46 | Improved |
| add positional clues to GA attention (ablation) | Polar + GRR + window attn · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 65.46 -> 66.57 | Improved |
| add geometric clues to GA attention (ablation) | Polar + GRR + window attn + pos clues · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 66.57 -> 67.84 | Improved |
| use axial, channel-wise or Swin attention for re-alignment instead of GRR (ablation) | Polar baseline · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 64.83 -> 64.92 (axial); 64.83 -> 65.22 (channel-wise); 64.83 -> 65.13 (Swin) | No clear effect |
| use uniform range partition instead of SID or LID (ablation) | Polar + GRR + GA · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 65.81 -> 67.84 (SID); 66.14 -> 67.84 (LID) | Improved |
| regress box targets in Cartesian instead of polar coordinates (ablation) | PARTNER · Waymo val vehicle ablation (12 epochs, L2 mAP) | L2 mAP: 67.52 -> 67.84 | Improved |
| switch CenterPoint from Cartesian to naive polar representation | CenterPoint · Waymo val (vehicle) | LEVEL 1 3D mAP: 75.58 -> 72.24 | Hurt |
| PARTNER polar representation at coarse 5x voxel size vs Cartesian | CenterPoint · Waymo val vehicle 5x (L2 mAPH) | L2 mAPH: 35.6 -> 50.3 | 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 4 of 4 shared results
- PV-RCNN — PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020) · better on 3 of 3 shared results
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 LiDAR 3D Object Detection (ICCV 2023 · similarity 0.97)
- PolarFormer: Multi-Camera 3D Object Detection with Polar Transformer (AAAI 2023 · similarity 0.90)
- PolarStream: Streaming Object Detection and Segmentation with Polar Pillars (NeurIPS 2021 · similarity 0.89)
- PolarNet: Learning to Optimize Polar Keypoints for Keypoint Based Object Detection (ICLR 2021 · similarity 0.87)
- PolarNeXt: Rethink Instance Segmentation with Polar Representation (CVPR 2025 · similarity 0.85)
- RGB-to-Polarization Estimation: A New Task and Benchmark Study (NeurIPS 2025 · similarity 0.84)
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.
