Research map / 3D vision / 3D generation
FB-BEV: BEV Representation from Forward-Backward View Transformations
Zhiqi Li, Zhiding Yu, Wenhai Wang, Anima Anandkumar, Tong Lu, Jose M. Alvarez
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 depth consistency (depth-aware weighting) to backward projection in FB-BEV (ablation) | FB-BEV · nuScenes val (R50, no temporal) | NDS: 0.397 -> 0.406 | Improved |
| add depth consistency to BEVFormer spatial cross-attention (ablation) | BEVFormer · nuScenes val (R50, no temporal) | NDS: 0.379 -> 0.387 NDS; mAP 0.297 -> 0.291 | Mixed |
| add depth consistency to BEVFormer-T with depth supervision (ablation) | BEVFormer-T · nuScenes val (R50, temporal, depth sup.) | NDS: 0.461 -> 0.472 | Improved |
| add depth supervision as auxiliary task to plain backward projection | BEVFormer-T · nuScenes val (R50, temporal) | NDS: 0.459 -> 0.461 | No clear effect |
| add FRPN foreground mask before backward projection (ablation) | FB-BEV · nuScenes val (R50, no temporal) | NDS / latency: NDS 0.400 -> 0.406; latency 3.4ms -> 2.6ms | Improved |
| add forward-backward projection over forward-only at large BEV size | BEVDet · nuScenes val (BEV 400x400) | NDS: 0.368 -> 0.406 | Improved |
| increase BEV size from 128 to 400 for forward-only BEVDet (ablation) | BEVDet · nuScenes val (BEV scale) | NDS: 0.370 -> 0.368 | No clear effect |
| replace BEVFormer/BEVDet VTM with FB-BEV (no temporal) | FB-BEV · nuScenes val (R50) | NDS: 0.382 -> 0.406 | Improved |
Later papers that compared with this method
1 later accepted paper reports FB-BEV as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Hybrid-grained Feature Aggregation with Coarse-to-fine Language Guidance for Self-supervised Monocular Depth Estimation (ICCV 2025 · better on 1 of 1 shared result)
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.
- BEVDepth* — BEVDepth: Acquisition of Reliable Depth for Multi-View 3D Object Detection (AAAI 2023) · better on 1 of 1 shared result
- BEVStereo — BEVStereo: Enhancing Depth Estimation in Multi-View 3D Object Detection with Temporal Stereo (AAAI 2023) · 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
- Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting (CVPR 2025 · similarity 0.86)
- M-BEV: Masked BEV Perception for Robust Autonomous Driving (AAAI 2024 · similarity 0.86)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.86)
- BerfScene: Bev-conditioned Equivariant Radiance Fields for Infinite 3D Scene Generation (CVPR 2024 · similarity 0.86)
- Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · similarity 0.85)
- VQ-Map: Bird's-Eye-View Map Layout Estimation in Tokenized Discrete Space via Vector Quantization (NeurIPS 2024 · similarity 0.84)
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.
