Research map / 3D vision / Depth and stereo
ARINBEV: Bird's-Eye View Layout Estimation with Conditional Autoregressive Model
Jiyong Kwag, Charles Toth, Alper Yilmaz
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 |
|---|---|---|---|
| use hybrid masking (random plus entropy-guided, p=0.5) instead of random masking only (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 62.9 -> 64.3 | Improved |
| use hybrid masking instead of pure entropy-guided masking (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 61.1 -> 64.3 | Improved |
| bilinear downsampling plus 3x3 conv instead of three 3x3 convs (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 63.9 -> 64.3 | Improved |
| bilinear downsampling plus 3x3 conv instead of single 8x8 stride-8 conv (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 64.0 -> 64.3 | Improved |
| add learnable conv after bilinear downsampling instead of pure bilinear (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 62.8 -> 64.3 | Improved |
| use structured class encoding instead of channel-wise binary encoding (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 61.0 -> 64.3 | Improved |
| class encoding scale 0.01 instead of 1.0 (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 61.2 -> 64.3 | Improved |
| class encoding fixed scale 0.01 instead of learnable scale (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 57.2 -> 64.3 | Improved |
| Gaussian prior sigma 0.5 instead of 0.3 (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 62.1 -> 64.3 | Improved |
| sample 3 steps instead of 1 (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 59.8 -> 64.3 | Improved |
| sample 4 steps instead of 3 (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 64.3 -> 64.3 | No clear effect |
| vary sigma range during training (0.40-0.70 vs 0.25-0.75) (ablation) | ARINBEV · nuScenes val (camera-only) | mean IoU: 63.9 -> 64.3 | No clear effect |
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.
- VQ-Map — VQ-Map: Bird's-Eye-View Map Layout Estimation in Tokenized Discrete Space via Vector Quantization (NeurIPS 2024) · better on 3 of 3 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
- VQ-Map: Bird's-Eye-View Map Layout Estimation in Tokenized Discrete Space via Vector Quantization (NeurIPS 2024 · similarity 0.89)
- MapPrior: Bird's-Eye View Map Layout Estimation with Generative Models (ICCV 2023 · similarity 0.88)
- Projecting Your View Attentively: Monocular Road Scene Layout Estimation via Cross-View Transformation (CVPR 2021 · similarity 0.87)
- Generative Lane Topology Reasoning via Autoregressive Model with Geometry Prior (ECCV 2026 · similarity 0.87)
- LaneSegNet: Map Learning with Lane Segment Perception for Autonomous Driving (ICLR 2024 · similarity 0.85)
- Map It Anywhere: Empowering BEV Map Prediction using Large-scale Public Datasets (NeurIPS 2024 · 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.
