Research map / Detection and segmentation / 3D detection and driving
Deep Continuous Fusion for Multi-Sensor 3D Object Detection
Ming Liang, Bin Yang, Shenlong Wang, Raquel Urtasun
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 image fusion via continuous fusion to the LIDAR BEV detector (ablation) | BEV detector · KITTI val Car 3D moderate | 3D AP (%): 65.90 -> 73.25 | Improved |
| add camera image fusion via continuous fusion to the LIDAR BEV detector (ablation) | BEV detector · KITTI val Car BEV moderate | BEV AP (%): 84.23 -> 87.34 | Improved |
| replace discrete fusion with continuous fusion (KNN pooling plus geometric feature) (ablation) | fusion detector · KITTI val Car 3D moderate | 3D AP (%): 67.79 -> 73.25 | Improved |
| add geometric feature to MLP input of continuous fusion (ablation) | continuous fusion · KITTI val Car 3D moderate | 3D AP (%): 70.09 -> 73.25 | Improved |
| add geometric feature to MLP input of continuous fusion (ablation) | continuous fusion · KITTI val Car BEV moderate | BEV AP (%): 86.37 -> 87.34 | Improved |
| add KNN pooling without geometric feature to discrete fusion (ablation) | fusion detector · KITTI val Car 3D moderate | 3D AP (%): 67.79 -> 70.09 | Improved |
| remove the distance threshold (d=10 to d=+inf) (ablation) | continuous fusion k=1 · KITTI val Car 3D moderate | 3D AP (%): 73.25 -> 72.39 | No clear effect |
| increase nearest neighbors from k=1 to k=5 (d=3) (ablation) | continuous fusion · KITTI val Car 3D moderate | 3D AP (%): 72.33 -> 71.20 | Hurt |
| add camera image fusion to BEV detector | BEV detector · TOR4D Vehicle | BEV AP0.7: 81.41 -> 83.89 | Improved |
| add camera image fusion to BEV detector | BEV detector · TOR4D Pedestrian | BEV AP0.3: 78.87 -> 82.32 | Improved |
| add camera image fusion to BEV detector | BEV detector · TOR4D Bicyclist | BEV AP0.3: 70.97 -> 74.08 | Improved |
| add camera image fusion to BEV detector | BEV detector · TOR4D Bicyclist | BEV AP0.5: 57.63 -> 59.83 | Improved |
Closest papers by meaning
- Multi-Task Multi-Sensor Fusion for 3D Object Detection (CVPR 2019 · similarity 0.90)
- 3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object Detection (ECCV 2020 · similarity 0.89)
- Towards Universal LiDAR-Based 3D Object Detection by Multi-Domain Knowledge Transfer (ICCV 2023 · similarity 0.88)
- Multi-Level Fusion Based 3D Object Detection From Monocular Images (CVPR 2018 · similarity 0.88)
- Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning (Student Abstract) (AAAI 2025 · similarity 0.87)
- PromptDet: A Lightweight 3D Object Detection Framework with LiDAR Prompts (AAAI 2025 · similarity 0.87)
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.
