Research map / Detection and segmentation / CNN features
Translating Images to Road Network: A Sequence-to-Sequence Perspective
Jiachen Lu, Ming Nie, Bozhou Zhang, Renyuan Peng, Xinyue Cai, Hang Xu, Feng Wen, Wei Zhang, Li Zhang
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 semi-autoregressive decoding instead of fully autoregressive | AR-RNTR · nuScenes multi-camera (default split) | Landmark F1: 54.1 -> 56.0 | Improved |
| use non-autoregressive masked decoding | AR-RNTR · nuScenes multi-camera (default split) | FPS: 0.1 -> 5.5 (47x) | Improved |
| add Topology-Inherited Training | AR-RNTR · nuScenes multi-camera (new split) | Reachability F1: 40.9 -> 49.8 | Improved |
| add Topology-Inherited Training | AR-RNTR · nuScenes multi-camera (new split) | Landmark F1: 36.5 -> 39.1 | Improved |
| add SD-Map prior | AR-RNTR + TIT · nuScenes multi-camera (new split) | Landmark F1: 39.1 -> 54.7 | Improved |
| add SD-Map prior | AR-RNTR + TIT · nuScenes multi-camera (new split) | Reachability F1: 49.8 -> 62.6 | Improved |
| add Intra-seq self-attention (ablation) | NAR-RNTR · nuScenes multi-camera | Reachability F1: 56.3 -> 66.0 | Improved |
| raise mask ratio from 50% to 90% (ablation) | NAR-RNTR · nuScenes multi-camera | Reachability F1: 51.3 -> 66.0 | Improved |
| use 3 inference iterations instead of 1 (ablation) | NAR-RNTR · nuScenes multi-camera | Reachability F1: 62.7 -> 66.0 | Improved |
| use 6 inference iterations instead of 3 (ablation) | NAR-RNTR · nuScenes multi-camera | Reachability F1: 66.0 -> 66.1 | No clear effect |
| add BEV augmentation (ablation) | NAR-RNTR · nuScenes multi-camera | Landmark F1: 57.5 -> 60.2 | Improved |
| add sequence noise padding (ablation) | NAR-RNTR · nuScenes multi-camera | Landmark F1: 58.6 -> 60.2 | Improved |
Closest papers by meaning
- Translating Images to Road Network: A Non-Autoregressive Sequence-to-Sequence Approach (ICCV 2023 · similarity 0.97)
- TD-Road: Top-Down Road Network Extraction with Holistic Graph Construction (ECCV 2022 · similarity 0.89)
- LineGraph2Road: Structural Graph Reasoning on Line Graphs for Road Network Extraction (ECCV 2026 · similarity 0.89)
- Topology Preserving Local Road Network Estimation From Single Onboard Camera Image (CVPR 2022 · similarity 0.88)
- Road Network Representation Learning with the Third Law of Geography (NeurIPS 2024 · similarity 0.87)
- RoadSceneBench: A Lightweight Benchmark for Mid-Level Road Scene Understanding (CVPR 2026 · 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.
