Research map / Detection and segmentation / 3D detection and driving
Voxel Field Fusion for 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 voxel field fusion | CenterPoint · nuScenes test | mAP: 60.3 -> 68.4; 8.1% mAP | Helped |
| Add voxel field fusion | CenterPoint · nuScenes test | NDS: 67.3 -> 72.4; 5.1% NDS | Helped |
| Add voxel field fusion | Voxel R-CNN · KITTI test, Car Hard | AP3D-R40: 77.06 -> 79.29; 2.2% AP | Helped |
| Use image-level flipping and rescaling aligned to point transformations rather than reproject alignment (ablation) | PV-RCNN with single fusion · KITTI val, Car Moderate | AP3D-R40: 82.50 -> 84.15 | Helped |
| Add the sample-added strategy to the mixed augmentor (ablation) | PV-RCNN + VFF · KITTI val, Car Moderate | AP3D-R40: 76.25 -> 82.53 | Helped |
| Add sample-static augmentation to sample-added augmentation (ablation) | PV-RCNN + VFF · KITTI val, Car Moderate | AP3D-R40: 82.53 -> 85.51 | Helped |
| Use learnable importance sampling instead of the strongest heuristic density sampling (ablation) | PV-RCNN + VFF · KITTI val, Car Moderate | AP3D-R40: 82.01 -> 82.10 | Helped |
| Extend local fusion to fusion along the whole ray (ablation) | PV-RCNN + VFF · KITTI val, Car Moderate | AP3D-R40: 82.12 -> 82.53 | Helped |
| Use Gaussian supervision with radius 1 instead of radius 0 without Gaussian supervision (ablation) | PV-RCNN + VFF · KITTI val, Car Moderate | AP3D-R40: 82.10 -> 82.53 | Helped |
| Expand Gaussian ray supervision radius from 1 to 2 (ablation) | PV-RCNN + VFF · KITTI val, Car Moderate | AP3D-R40: 82.53 -> 81.75 | Hurt |
| Apply fusion at Stage-1 rather than the no-fusion baseline (ablation) | PV-RCNN · KITTI val, Car Moderate | AP3D-R40: 79.19 -> 82.53; 3.34% AP | Helped |
| Move fusion from Stage-1 to Stage-3 (ablation) | PV-RCNN · KITTI val, Car Moderate | AP3D-R40: 82.53 -> 80.27 | Hurt |
Closest papers by meaning
- Unifying Voxel-based Representation with Transformer for 3D Object Detection (NeurIPS 2022 · similarity 0.90)
- VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking (CVPR 2023 · similarity 0.88)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.88)
- GO-N3RDet: Geometry Optimized NeRF-enhanced 3D Object Detector (CVPR 2025 · similarity 0.87)
- VeXKD: The Versatile Integration of Cross-Modal Fusion and Knowledge Distillation for 3D Perception (NeurIPS 2024 · similarity 0.85)
- SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · similarity 0.85)
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.