Research map / Detection and segmentation / 3D detection and driving
DeepFusion: Lidar-Camera Deep Fusion for Multi-Modal 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 camera features with DeepFusion to PointPillars | PointPillars · Waymo Open Dataset validation, Vehicle L2 | APH: 48.1 -> 54.8 (+6.7) | Helped |
| Add camera features with DeepFusion to CenterPoint | CenterPoint · Waymo Open Dataset validation, Vehicle L2 | APH: 54.3 -> 63.2 (+8.9) | Helped |
| Add camera features with DeepFusion to 3D-MAN | 3D-MAN · Waymo Open Dataset validation, Vehicle L2 | APH: 52.2 -> 58.4 (+6.2) | Helped |
| Add InverseAug to deep feature fusion (ablation) | DeepFusion (3D-MAN++) · Waymo Open Dataset validation, Pedestrian L2 | APH: 63.5 -> 67.0 | Helped |
| Add LearnableAlign to deep feature fusion (ablation) | DeepFusion (3D-MAN++) · Waymo Open Dataset validation, Pedestrian L2 | APH: 66.4 -> 67.0 | Helped |
| Fuse deep features with DeepFusion rather than LateFusion | 3D-MAN++ multimodal detector · Waymo Open Dataset validation, Pedestrian L2 | APH: 66.5 -> 67.0 (+0.5) | Helped |
| Use multimodal fusion with increasing maximum RandomRotation (ablation) | 3D-MAN++ · Waymo Open Dataset validation, Pedestrian L1 | AP gain over single-modal: +2.6 -> +0.9 -> +0.8 -> +0.4 | Hurt |
| Use multimodal fusion with increasing RandomFlip probability (ablation) | 3D-MAN++ · Waymo Open Dataset validation, Pedestrian L1 | AP gain over single-modal: +2.3 -> +0.10 -> +0.03 | Hurt |
| Add the NAS-designed feature encoder and replace ReLU with SILU (ablation) | 3D-MAN · Waymo Open Dataset validation, Pedestrian L2 | APH: 52.2 -> 63.0 | Helped |
| Use DeepFusion instead of single-modal input under Laser Noise | 3D-MAN++ · Waymo Open Dataset validation, Pedestrian L2, Laser Noise | APH: 52.3 -> 66.8 | Helped |
| Add DeepFusion to the lidar-only model on Kirkland OOD data | 3D-MAN++ · Waymo Open Dataset Kirkland validation, Pedestrian L2 | APH: 27.6 -> 34.6 (+8.0) | Helped |
Closest papers by meaning
- MSMDFusion: Fusing LiDAR and Camera at Multiple Scales With Multi-Depth Seeds for 3D Object Detection (CVPR 2023 · similarity 0.92)
- BEVFusion: A Simple and Robust LiDAR-Camera Fusion Framework (NeurIPS 2022 · similarity 0.92)
- TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection With Transformers (CVPR 2022 · similarity 0.92)
- GAFusion: Adaptive Fusing LiDAR and Camera with Multiple Guidance for 3D Object Detection (CVPR 2024 · similarity 0.90)
- SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · similarity 0.89)
- PromptDet: A Lightweight 3D Object Detection Framework with LiDAR Prompts (AAAI 2025 · 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.