Research map / Detection and segmentation / 3D detection and driving
Center Focusing Network for Real-Time LiDAR Panoptic Segmentation
Xiaoyan Li, Gang Zhang, Boyue Wang, Yongli Hu, Baocai Yin
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 CFFE with center feature generation to CPGNet (ablation) | CFNet · SemanticKITTI validation | PQ: 60.8 -> 62.7 | Improved |
| Add intermediate supervision and FEM without center feature generation (ablation) | CPGNet with CDM · SemanticKITTI validation | PQ: 60.5 -> 60.8 | Improved |
| Replace BEV center heatmap with CDM (ablation) | PolarNet · SemanticKITTI validation | PQ: 60.4 -> 60.6 | Improved |
| Replace DBSCAN, HDBSCAN, or MeanShift with CDM (ablation) | CPGNet · SemanticKITTI validation | PQ: +0.6~+2.2 PQ | Improved |
| Replace DBSCAN, HDBSCAN, or MeanShift with CDM (ablation) | CPGNet · SemanticKITTI validation | RT (ms): -21.5 ms~-81.2 ms | Improved |
| Use wider FEM receptive fields instead of setting all dilation coefficients to 1 (ablation) | CFNet with CPGNet · SemanticKITTI validation | PQ: 62.2 -> 62.7 | Improved |
| Use ground-truth offsets instead of predicted offsets for CFG (ablation) | CFNet with CPGNet · SemanticKITTI validation | PQ: 62.7 -> 65.5 | Improved |
| Use predicted CFG center offsets rather than ground-truth center offsets (ablation) | CFNet with CPGNet · SemanticKITTI validation | things center offset error: 1.13 vs 1.22 | Hurt |
| Use a class-independent CDM distance threshold (ablation) | CFNet · SemanticKITTI validation | PQ: car: 95.2 at d=1.2; person: 85.5 at d=0.8 | Mixed |
| Use CFNet instead of Panoptic-PHNet | LiDAR panoptic segmentation · SemanticKITTI test | PQ: 61.5 -> 63.4 | Improved |
| Use CFNet instead of Panoptic-PHNet | LiDAR panoptic segmentation · nuScenes validation | PQ: 74.7 -> 75.1 | Improved |
| Use CFNet instead of Panoptic-PHNet | LiDAR panoptic segmentation · nuScenes test | PQ: 80.1 -> 79.4 | Hurt |
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.
- Panoptic-PHNet — Panoptic-PHNet: Towards Real-Time and High-Precision LiDAR Panoptic Segmentation via Clustering Pseudo Heatmap (CVPR 2022) · better on 2 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
- Panoptic-PHNet: Towards Real-Time and High-Precision LiDAR Panoptic Segmentation via Clustering Pseudo Heatmap (CVPR 2022 · similarity 0.89)
- Panoptic-PolarNet: Proposal-Free LiDAR Point Cloud Panoptic Segmentation (CVPR 2021 · similarity 0.89)
- LiDAR-Based Panoptic Segmentation via Dynamic Shifting Network (CVPR 2021 · similarity 0.89)
- Sparse Cross-Scale Attention Network for Efficient LiDAR Panoptic Segmentation (AAAI 2022 · similarity 0.88)
- Real-Time Panoptic Segmentation From Dense Detections (CVPR 2020 · similarity 0.87)
- How Do Images Align and Complement LiDAR? Towards a Harmonized Multi-modal 3D Panoptic Segmentation (ICML 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.
