Research map / Graph learning / Graph self-supervised learning
LaneDiffusion: Improving Centerline Graph Learning via Prior Injected BEV Feature Generation
Zijie Wang, Weiming Zhang, Wei Zhang, Xiao Tan, Hongxing Liu, Yaowei Wang, Guanbin Li
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 |
|---|---|---|---|
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · nuScenes point-level | GEO F1: 54.7 -> 58.9 | Improved |
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · nuScenes point-level | TOPO F1: 42.2 -> 46.8 | Improved |
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · nuScenes point-level | APLS: 30.7 -> 37.1 | Improved |
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · nuScenes segment-level | mAP_cf: 35.2 -> 41.6 | Improved |
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · nuScenes segment-level | TOP_ll: 1.3 -> 3.4 | Improved |
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · Argoverse2 val | TOPO F1: 44.5 -> 47.0 | Improved |
| add LaneDiffusion BEV prior diffusion to CGNet | CGNet · Argoverse2 val | TOP_ll: 0.5 -> 1.8 | Improved |
| use T=15 sampling steps instead of T=5 (ablation) | LaneDiffusion · nuScenes | APLS: 33.1 -> 37.1 | Improved |
| use T=30 sampling steps instead of T=15 (ablation) | LaneDiffusion · nuScenes | TOPO F1: 46.8 -> 46.0 | Hurt |
| train LPDM with Overall paradigm instead of Sampling (T=5) (ablation) | LaneDiffusion · nuScenes | TOPO F1: 44.8 -> 45.6 | Improved |
| add feature refinement (Concat&ED) after generation (ablation) | LaneDiffusion · nuScenes | TOPO F1: 43.9 -> 46.8 | Improved |
| use encoder-decoder instead of 1x1 conv for refinement (concat) (ablation) | LaneDiffusion · nuScenes | APLS: 35.7 -> 37.1 | Improved |
Closest papers by meaning
- LaneDiffusion++: Feature-Level Generative Priors for Structured Lane Graph Learning (TPAMI 2026 · similarity 0.93)
- LaneSegNet: Map Learning with Lane Segment Perception for Autonomous Driving (ICLR 2024 · similarity 0.91)
- LaneGraph2Seq: Lane Topology Extraction with Language Model via Vertex-Edge Encoding and Connectivity Enhancement (AAAI 2024 · similarity 0.88)
- Controllable Road Marking Generation (NeurIPS 2026 · similarity 0.88)
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.87)
- Bezier Everywhere All at Once: Learning Drivable Lanes as Bezier Graphs (CVPR 2024 · 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.
