Research map / Detection and segmentation / 3D detection and driving
Fully Sparse Fusion for 3D Object Detection
Yingyan Li, Lue Fan, Yang Liu, Zehao Huang, Yuntao Chen, Naiyan Wang, Zhaoxiang 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 camera instances to LiDAR-instance-only model (ablation) | FSF · nuScenes val | mAP: 62.5 -> 68.6 | Improved |
| add LiDAR instances to camera-instance-only model (ablation) | FSF · nuScenes val | mAP: 63.1 -> 68.6 | Improved |
| add instance shape alignment (ablation) | FSF · nuScenes val | mAP: 68.6 -> 70.4 | Improved |
| add instance feature interaction module (ablation) | FSF · nuScenes val | NDS: 72.1 -> 72.7 | Improved |
| add second-stage camera-based assignment to LiDAR-based assignment (ablation) | FSF · nuScenes val | mAP: 69.0 -> 70.4 | Improved |
| use LiDAR-based instead of point-in-box assignment (ablation) | FSF · nuScenes val | mAP: 68.7 -> 69.0 | Improved |
| use Hungarian assignment for paired camera and LiDAR instances (ablation) | FSF · nuScenes val | mAP: Not converged | Hurt |
| use BEV-distance second stage instead of camera-based (ablation) | FSF · nuScenes val | mAP: 70.4 -> 69.2 | Hurt |
| replace HTC with weaker Mask RCNN as 2D model (ablation) | FSF · nuScenes val | mAP: 70.4 -> 69.4 | Hurt |
| degrade 2D masks to boxes (ablation) | FSF · nuScenes val | mAP: 70.4 -> 69.8 | Hurt |
| add image features from a trained image backbone (ablation) | FSF · nuScenes val | mAP: 70.4 -> 70.7 | No clear effect |
| instance-level fusion instead of score painting on FSD | FSF vs FSD+Score Painting · nuScenes val | mAP: 66.9 -> 70.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DeepInteraction — DeepInteraction: 3D Object Detection via Modality Interaction (NeurIPS 2022) · better on 3 of 4 shared results
- VoxelNeXt — VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking (CVPR 2023) · better on 2 of 2 shared results
- TransFusion — TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection With Transformers (CVPR 2022) · better on 1 of 1 shared result
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
- Fully Sparse 3D Object Detection (NeurIPS 2022 · similarity 0.94)
- SparseLIF: High-Performance Sparse LiDAR-Camera Fusion for 3D Object Detection (ECCV 2024 · similarity 0.93)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.90)
- Far3D: Expanding the Horizon for Surround-View 3D Object Detection (AAAI 2024 · similarity 0.89)
- SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · similarity 0.89)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 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.
