Research map / Detection and segmentation / 3D detection and driving
HVNet: Hybrid Voxel Network for LiDAR Based 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 |
|---|---|---|---|
| Add attentive VFE to the single-scale baseline (ablation) | HVNet · KITTI validation, BEV Moderate | mAP: 71.7 -> 73.76 | Helped |
| Add FPN to the attention model (ablation) | HVNet · KITTI validation, BEV Moderate | mAP: 73.76 -> 74.34 | Helped |
| Use FFPN instead of FPN (ablation) | HVNet · KITTI validation, BEV Moderate | mAP: 73.76 -> 75.18 | Helped |
| Set focal-loss α to 0.75 instead of 0.25 (ablation) | HVNet with FFPN · KITTI validation, BEV Moderate | mAP: 75.18 -> 75.71 | Helped |
| Increase feature voxel scales to two (ablation) | HVNet · KITTI validation, BEV Moderate | mAP: 75.71 -> 76.88 | Helped |
| Add a second projection scale (ablation) | HVNet with two feature scales · KITTI validation, BEV Moderate | mAP: 76.88 -> 77.88 | Helped |
| Increase feature and projection scales to three (ablation) | HVNet with two feature and projection scales · KITTI validation, BEV Moderate | mAP: 77.88 -> 78.59 | Helped |
| Regress corner offsets rather than box pose parameters (ablation) | HVNet · KITTI validation, Car BEV Moderate | AP: 87.67 -> 88.90 | Helped |
| Add an IoU prediction branch and multiply its prediction with classification scores (ablation) | HVNet · KITTI test | AP | Mixed |
| Use IoU directly as the loss (ablation) | HVNet · KITTI validation and test | AP | Hurt |
| Use uncertainty-based learned weights for loss terms (ablation) | HVNet · KITTI | AP | No clear effect |
| Use hybrid voxel scales and index-based implementation instead of fixed-buffer VFE (ablation) | HVNet feature extractor · Inference time | time: 34ms -> 32ms | Helped |
Closest papers by meaning
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.94)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.92)
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.92)
- VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking (CVPR 2023 · similarity 0.92)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.91)
- Voxel Mamba: Group-Free State Space Models for Point Cloud based 3D Object Detection (NeurIPS 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.