Research map / Detection and segmentation / 3D detection and driving
GaussianFusion: Unified 3D Gaussian Representation for Multi-Modal Fusion Perception
Xiao Zhao, Chang Liu, Mingxu Zhu, Zheyuan Zhang, Linna Song, Qingliang Luo, Chufan Guo, Kuifeng Su
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 |
|---|---|---|---|
| replace BEV grid fusion with unified 3D Gaussian fusion | BEVFusion · nuScenes val detection | NDS: 71.4 -> 74.0 | Improved |
| Gaussian representation instead of BEV grid on Waymo | BEVFusion · Waymo Open Dataset | mAP-L1: 82.72 -> 86.23 | Improved |
| Gaussian representation at low resolution (100x100) | BEVFusion · nuScenes val | NDS: 70.5 -> 73.1 | Improved |
| increase BEV size 100x100 to 400x400 (ablation) | BEVFusion · nuScenes val | NDS: 70.5 -> 72.7 | Improved |
| increase BEV size 100x100 to 400x400 (ablation) | GaussianFusion · nuScenes val | NDS: 73.1 -> 74.4 | No clear effect |
| forward-projection Gaussian init vs random init (ablation) | GaussianFusion · nuScenes val | NDS: 71.2 -> 74.0 | Improved |
| forward-projection init vs backward projection (ablation) | GaussianFusion · nuScenes val | NDS: 72.4 -> 74.0 | Improved |
| forward-projection init vs Lidar projection init (ablation) | GaussianFusion · nuScenes val | NDS: 73.6 -> 74.0 | No clear effect |
| use shared Gaussian encoder instead of separate (ablation) | GaussianFusion · nuScenes val | mAP: 71.0 -> 71.7 | Improved |
| add Gaussian priors to deformable attention (ablation) | GaussianFusion · nuScenes val | NDS: 73.6 -> 74.0 | Improved |
| encode Gaussian properties as positional encoding in query (ablation) | GaussianFusion · nuScenes val | mAP: 71.2 -> 71.7 | Improved |
| predict property offsets instead of new properties (ablation) | GaussianFusion · nuScenes val | mAP: 70.8 -> 71.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- UniTR — UniTR: A Unified and Efficient Multi-Modal Transformer for Bird's-Eye-View Representation (ICCV 2023) · better on 4 of 4 shared results
- MetaBEV — MetaBEV: Solving Sensor Failures for 3D Detection and Map Segmentation (ICCV 2023) · better on 2 of 2 shared results
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
- Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting (CVPR 2025 · similarity 0.88)
- VeXKD: The Versatile Integration of Cross-Modal Fusion and Knowledge Distillation for 3D Perception (NeurIPS 2024 · similarity 0.87)
- GAFusion: Adaptive Fusing LiDAR and Camera with Multiple Guidance for 3D Object Detection (CVPR 2024 · similarity 0.86)
- IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · similarity 0.86)
- GaussianFusion: Gaussian-Based Multi-Sensor Fusion for End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.86)
- MamBEV: Enabling State Space Models to Learn Birds-Eye-View Representations (ICLR 2025 · 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.
