Research map / 3D vision / 3D generation
Towards Viewpoint Robustness in Bird's Eye View Segmentation
Tzofi Klinghoffer, Jonah Philion, Wenzheng Chen, Or Litany, Zan Gojcic, Jungseock Joo, Ramesh Raskar, Sanja Fidler, 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 |
|---|---|---|---|
| Train BEV model with 25-50% of data transformed to target viewpoint by NVS | CVT · DRIVE Sim, pitch -10 deg | IoU: 0.014 -> 0.165 | Improved |
| Train BEV model with NVS-transformed data for +0.8 m height target | CVT · DRIVE Sim, height 0.8 m | IoU: 0.003 -> 0.214 | Improved |
| Pass train extrinsics at inference (Source*) instead of target extrinsics | CVT source model · DRIVE Sim, pitch -10 deg | IoU: 0.014 -> 0.078 | Improved |
| Randomly rotate extrinsics and 3D box labels during training (Extr Aug) | CVT · DRIVE Sim, pitch +5 deg | IoU: 0.016 -> 0.028 | Improved |
| Extrinsic augmentation at source viewpoint | CVT · DRIVE Sim, source viewpoint | IoU: 0.170 -> 0.155 | Hurt |
| Transform 100% of training data instead of 25-50% (ablation) | NVS augmentation · DRIVE Sim target rigs | IoU | Hurt |
| Increase transformed fraction up to 50% (ablation) | NVS augmentation · DRIVE Sim target rigs | IoU | Improved |
| Add SSIM loss, minimum loss and automasking to Worldsheet (ablation) | Worldsheet NVS · AV NVS, 1K images | Depth L1: 0.00763 -> 0.00707 | Improved |
| Add lidar depth supervision (ablation) | Worldsheet NVS · AV NVS, 1K images | Depth L1: 0.00707 -> 0.00657 | Improved |
| Train on multiple viewpoints rather than only the target view | CVT · DRIVE Sim, mean over views | IoU: 0.19 mean test IoU (0.206 train views, 0.178 other views) | Improved |
| Change camera viewpoint at inference for a source-trained model (CARLA analysis) | LSS and CVT · CARLA | IoU | Hurt |
Closest papers by meaning
- Benchmarking and Improving Bird’s Eye View Perception Robustness in Autonomous Driving (TPAMI 2025 · similarity 0.93)
- A Black-Box Evaluation Framework for Semantic Robustness in Bird’s Eye View Detection (AAAI 2025 · similarity 0.91)
- Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · similarity 0.90)
- M-BEV: Masked BEV Perception for Robust Autonomous Driving (AAAI 2024 · similarity 0.89)
- Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting (CVPR 2025 · similarity 0.88)
- Predictive Uncertainty Quantification for Bird's Eye View Segmentation: A Benchmark and Novel Loss Function (ICLR 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.
