Research map / Graph learning / Graph self-supervised learning
SeqGrowGraph: Learning Lane Topology as a Chain of Graph Expansions
Mengwei Xie, Shuang Zeng, Xinyuan Chang, Xinran Liu, Zheng Pan, Mu Xu, Xing Wei
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 DFS node order instead of Center order (ablation) | SeqGrowGraph · nuScenes | Landmark F1: 52.4 -> 56.4 | Improved |
| use DFS node order instead of Coord order (ablation) | SeqGrowGraph · nuScenes | Landmark F1: 53.7 -> 56.4 | Improved |
| use DFS node order instead of BFS order (ablation) | SeqGrowGraph · nuScenes | Landmark F1: 55.1 -> 56.4 | Improved |
| use DFS instead of Center order (ablation) | SeqGrowGraph · nuScenes | Reachability F1: 63.7 -> 67.8 | Improved |
| use DFS instead of BFS (ablation) | SeqGrowGraph · nuScenes | Reachability F1: 67.7 -> 67.8 | Improved |
| raise loss weight of position tokens from 1 to 2 (ablation) | SeqGrowGraph · nuScenes | Landmark F1: 54.7 -> 56.4 | Improved |
| raise loss weight of position tokens from 2 to 3 (ablation) | SeqGrowGraph · nuScenes | Landmark F1: 56.4 -> 55.3 | Hurt |
| raise position loss weight from 2 to 4 (ablation) | SeqGrowGraph · nuScenes | Reachability F1: 67.8 -> 67.5 | No clear effect |
| re-segment centerlines every 40m instead of original split (ablation) | SeqGrowGraph · nuScenes | Junction Reachability F1: 62.2 -> 64.6 | Improved |
| re-segment centerlines every 20m instead of original split (ablation) | SeqGrowGraph · nuScenes | Junction Landmark F1: 59.5 -> 60.2 | Improved |
| re-segment centerlines every 30m instead of original split (ablation) | SeqGrowGraph · nuScenes | Junction Landmark F1: 59.5 -> 59.6 | No clear effect |
| model lane graph as sequential expansion instead of LaneGraph2Seq on nuScenes default | SeqGrowGraph vs LaneGraph2Seq* · nuScenes (default split) | Reachability F1: 46.2 -> 67.8 | Improved |
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.
- LaneGraph2Seq* — LaneGraph2Seq: Lane Topology Extraction with Language Model via Vertex-Edge Encoding and Connectivity Enhancement (AAAI 2024) · better on 8 of 8 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
- LaneGraph2Seq: Lane Topology Extraction with Language Model via Vertex-Edge Encoding and Connectivity Enhancement (AAAI 2024 · similarity 0.93)
- Bezier Everywhere All at Once: Learning Drivable Lanes as Bezier Graphs (CVPR 2024 · similarity 0.89)
- Topo2Seq: Enhanced Topology Reasoning via Topology Sequence Learning (AAAI 2025 · similarity 0.89)
- LaneSegNet: Map Learning with Lane Segment Perception for Autonomous Driving (ICLR 2024 · similarity 0.89)
- T2SG: Traffic Topology Scene Graph for Topology Reasoning in Autonomous Driving (CVPR 2025 · similarity 0.89)
- TopoLogic: An Interpretable Pipeline for Lane Topology Reasoning on Driving Scenes (NeurIPS 2024 · 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.
