Research map / Detection and segmentation / 3D detection and driving
UniTR: A Unified and Efficient Multi-Modal Transformer for Bird's-Eye-View Representation
Haiyang Wang, Hao Tang, Shaoshuai Shi, Aoxue Li, Zhenguo Li, Bernt Schiele, Liwei Wang
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 2D perspective-space cross-modal fusion to lidar-only model (ablation) | UniTR · nuScenes val | NDS: 70.5 -> 72.5 | Improved |
| add 3D geometric-space cross-modal fusion to lidar-only model (ablation) | UniTR · nuScenes val | NDS: 70.5 -> 72.0 | Improved |
| combine 2D and 3D fusion spaces (ablation) | UniTR · nuScenes val | NDS / mAP: 70.5 -> 73.1 NDS, 65.9 -> 70.0 mAP | Improved |
| add BEV LSS fusion step (ablation) | UniTR · nuScenes val | NDS: 73.1 -> 73.3 | Improved |
| parallel weight-sharing intra-modal block instead of serial modality-specific encoders (ablation) | UniTR · nuScenes val | latency (ms): 51 -> 33 | Improved |
| parallel weight-sharing instead of serial (ablation) | UniTR · nuScenes val | NDS: 72.2 -> 72.4 | Improved |
| place inter-modal blocks before intra-modal block (ablation) | UniTR · nuScenes val | NDS: 73.3 -> 72.9 | Hurt |
| order 3D fusion before 2D fusion (ablation) | UniTR · nuScenes val | NDS: 73.3 -> 73.0 | Hurt |
| parallel instead of serial processing under low-beam LiDAR (1-beam) (ablation) | UniTR · nuScenes val 1-beam LiDAR + camera | NDS: 57.6 -> 59.5 | Improved |
| parallel instead of serial processing with 32-beam LiDAR (ablation) | UniTR · nuScenes val 32-beam LiDAR + camera | NDS: 73.2 -> 73.3 | No clear effect |
| replace modality-specific encoders and late fusion with unified UniTR backbone | BEVFusion framework · nuScenes val 3D detection | NDS / latency: 71.4 -> 73.1 NDS; 130.5 -> 88.7 ms | Improved |
| UniTR lidar-only variant versus prior lidar methods on BEV segmentation | UniTR · nuScenes val BEV map segmentation | mean IoU: 48.6 -> 71.2 (vs CenterPoint) | Improved |
Later papers that compared with this method
6 later accepted papers report UniTR as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- GaussianFusion: Unified 3D Gaussian Representation for Multi-Modal Fusion Perception (ICLR 2026 · better on 4 of 4 shared results)
- BEVDilation: LiDAR-Centric Multi-Modal Fusion for 3D Object Detection (AAAI 2026 · better on 4 of 4 shared results)
- EVT: Efficient View Transformation for Multi-Modal 3D Object Detection (ICCV 2025 · better on 4 of 4 shared results)
- Height-Fidelity Dense Global Fusion for Multi-modal 3D Object Detection (ICCV 2025 · better on 4 of 4 shared results)
- DeepInteraction++: Multi-Modality Interaction for Autonomous Driving (TPAMI 2025 · better on 1 of 2 shared results)
- Detecting As Labeling: Rethinking LiDAR-camera Fusion in 3D Object Detection (ECCV 2024 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DeepInteraction — DeepInteraction: 3D Object Detection via Modality Interaction (NeurIPS 2022) · 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
- UniM2AE: Multi-modal Masked Autoencoders with Unified 3D Representation for 3D Perception in Autonomous Driving (ECCV 2024 · similarity 0.88)
- UniFusion: Unified Multi-View Fusion Transformer for Spatial-Temporal Representation in Bird's-Eye-View (ICCV 2023 · similarity 0.87)
- BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (ECCV 2022 · similarity 0.87)
- BEVFormer: Learning Bird’s-Eye-View Representation From LiDAR-Camera via Spatiotemporal Transformers (TPAMI 2024 · similarity 0.87)
- SGFormer: Semantic-Geometry Fusion Transformer for Multi-modal 3D Panoptic Segmentation (AAAI 2025 · similarity 0.86)
- VeXKD: The Versatile Integration of Cross-Modal Fusion and Knowledge Distillation for 3D Perception (NeurIPS 2024 · similarity 0.86)
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.
