Research map / Point clouds / Segmentation and detection
HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| combine voxel-based features with point-based features (ablation) | PointPillars-style voxel representation on KITTI validation · KITTI validation, car easy/moderate/hard | mAP (%): 1.31, 1.53 and 1.94 | Helped |
| use a memory module to replace point-feature extraction at test time (ablation) | HVPR with voxel-point features · KITTI validation, car | runtime (Hz): 6.8 -> 37.9 | Helped |
| use a memory module (ablation) | HVPR with voxel-point features · KITTI validation, car easy/moderate/hard | mAP (%): 88.97 -> 87.96; 80.35 -> 80.44; 77.28 -> 77.92 | No clear effect |
| refine multi-scale features with AMFM spatial attention (ablation) | HVPR with memory · KITTI validation, car easy/moderate/hard | mAP (%): 87.96 -> 89.51; 80.44 -> 80.86; 77.92 -> 78.22 | Helped |
| add explicit 3D scale features to AMFM (ablation) | HVPR with AMFM · KITTI validation, car easy/moderate/hard | mAP (%): 1.63, 1.19 and 1.27 | Helped |
| increase retrieved memory items from K=5 to K=20 (ablation) | HVPR memory · KITTI validation, car moderate; T=1000 | mAP (%): 80.22 -> 81.32 | Helped |
| increase retrieved memory items from K=20 to K=30 (ablation) | HVPR memory · KITTI validation, car moderate; T=1000 | mAP (%): 81.32 -> 80.82 | Hurt |
| increase memory size from T=1000 to T=3000 (ablation) | HVPR memory · KITTI validation, car moderate; K=20 | mAP (%): 81.32 -> 81.56 | Mixed |
| use HVPR instead of Associate-3D | voxel pseudo-image detection · KITTI test, car hard | mAP (%): 70.53 -> 73.04 | Helped |
| use HVPR instead of PointPillars | voxel pseudo-image detection · KITTI test, car easy/moderate/hard | mAP (%): 82.58 -> 86.38; 74.31 -> 77.92; 68.99 -> 73.04 | Helped |
| use HVPR instead of PointPillars | voxel pseudo-image detection · KITTI test | runtime (Hz): 42.4 -> 36.1 | Hurt |
Closest papers by meaning
- Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection (AAAI 2021 · similarity 0.93)
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.92)
- HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection (CVPR 2020 · similarity 0.92)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.89)
- Voxel Mamba: Group-Free State Space Models for Point Cloud based 3D Object Detection (NeurIPS 2024 · similarity 0.89)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.89)
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.