基于GNN-Transformer模型的车道线检测方法
贾子厚,罗佳,郑利锋,张艳岗,刘正阳,刘奔飞

Lane line detection method based on GNN-Transformer model
Zihou JIA,Jia LUO,Lifeng ZHENG,Yangang ZHANG,Zhengyang LIU,Benfei LIU
表2 各方法在CULane测试集的性能对比
Table 2 Performance comparison of each method on CULane test set
方法

F1@50/%

总分数

F1@50/%

十字路口

假阳性数

正常拥挤炫光阴影无线箭头曲线夜间
UFSA68.487.766.058.462.840.281.057.962.11 743
PINet74.490.372.366.368.449.883.765.666.71 427
STLNet73.691.870.265.969.348.885.367.569.21 887
E-CLRNet79.893.678.374.880.953.890.474.275.21 198
LaneATT75.191.272.765.868.049.187.863.768.61 020
本文方法74.394.471.868.868.648.185.665.667.51 450