AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

LineGraph2Road: Structural Graph Reasoning on Line Graphs for Road Network Extraction

Zhengyang Wei, Renzhi Jing, Yiyi He, Jenny Suckale

ECCV 2026 · Graph learning › Knowledge and temporal graphs · approach: Graph reasoning for visual recognition · 1 citations (OpenAlex)

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 changedWhereMeasuredEffect
add pretrained SAM weights to the encoder (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1 / APLS: 77.08 -> 82.37 F1; 61.63 -> 68.88 APLSImproved
add SAM image features for edge embedding (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1: 76.12 -> 82.37Improved
add SAM image features for edge embedding (ablation)LineGraph2Road · City-scale (ablation, non-overpass)APLS: 69.17 -> 68.88Hurt
add Graph Transformer over fully connected layer (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1 / APLS: 76.12 -> 82.37 F1; 55.84 -> 68.88 APLSImproved
use Graph Transformer instead of 3-layer MLP (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1 / APLS: 74.20 -> 82.37 F1; 52.20 -> 68.88 APLSImproved
increase sampled features per edge from 4 to 8 (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1 / APLS: 82.37 -> 82.34 F1; 68.88 -> 68.91 APLSNo clear effect
reduce sampled features per edge from 4 to 2 (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1 / APLS: 82.37 -> 82.25 F1; 68.88 -> 68.11 APLSNo clear effect
apply Graph Transformer on line graph instead of original graph (ablation)LineGraph2Road · City-scale (ablation, non-overpass)APLS: 65.72 -> 68.11Improved
use line-graph Graph Transformer instead of RelationFormer decoder (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1: 78.82 -> 82.25Improved
use line-graph Graph Transformer instead of Any2Graph decoder (ablation)LineGraph2Road · City-scale (ablation, non-overpass)TOPO F1: 77.53 -> 82.25Improved
add coupled NMS instead of standard NMS (ablation)LineGraph2Road · SpaceNetTOPO F1 / APLS: 83.46 -> 84.24 F1; 73.05 -> 73.94 APLSImproved
add coupled NMS instead of standard NMS (with overpass head) (ablation)LineGraph2Road w/ Overpass · City-scaleTOPO F1 / APLS: 82.11 -> 83.77 F1; 68.85 -> 70.40 APLSImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy