Research map / Reinforcement learning / Visual and representation RL
Mastering Spatial Graph Prediction of Road Networks
Anagnostidis Sotiris, Aurelien Lucchi, Thomas Hofmann
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 with reward and value estimates and tree search instead of cross-entropy autoregressive ARM | ARM · SpaceNet (ground truth key points) | APLS: 0.743 -> 0.778 | Improved |
| generate edges in arbitrary order instead of fixed autoregressive order | ARM · SpaceNet (ground truth key points) | bits per edge, random order: 30.84 -> 4.565 | Improved |
| use ours instead of ARM on autoregressive-order perplexity | ARM · SpaceNet (ground truth key points) | bits per edge, autoregressive order: 0.733 -> 4.636 | Hurt |
| attend to ViT per-patch image features in the two transformer modules (ablation) | Our method · SpaceNet | topological metrics | No clear effect |
| apply model to DeepGlobe without fine-tuning to refine baseline predictions | Our method · DeepGlobe | APLS: 0.6525 -> 0.7398 | Improved |
| use smaller simulation depth during training (ablation) | Our method · SpaceNet | topological metrics | No clear effect |
| use our method instead of Orientation on SpaceNet | Orientation · SpaceNet | CCQ quality: 0.6862 -> 0.6632 | Mixed |
Closest papers by meaning
- Using Reinforcement Learning to Iteratively Construct Road Networks from Satellite Images and GPS Data (AAAI 2024 · similarity 0.89)
- LineGraph2Road: Structural Graph Reasoning on Line Graphs for Road Network Extraction (ECCV 2026 · similarity 0.89)
- Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · similarity 0.87)
- Road Network Representation Learning with the Third Law of Geography (NeurIPS 2024 · similarity 0.86)
- Designing Specialized Two-Dimensional Graph Spectral Filters for Spatial-Temporal Graph Modeling (AAAI 2025 · similarity 0.86)
- Patched Line Segment Learning for Vector Road Mapping (AAAI 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.
