基于内容引导注意力的车道线检测网络

刘登峰,郭文静,陈世海

Content-guided attention-based lane detection network

Dengfeng LIU,Wenjing GUO,Shihai CHEN

表 3 CGANet在Tusimple上的实验结果

Tab.3 CGANet’s experimental results on TuSimple dataset

方法 基线网络 F1/% Acc/% PFP/% PFN/% SCNN[8] VGG16 94.97 93.12 7.17 2.20 UFLD[2] ResNet18 85.87 93.82 20.05 8.92 ResNet34 86.02 92.86 19.91 8.75 LaneATT[15] ResNet18 95.71 92.10 4.56 8.01 ResNet34 95.77 92.63 4.53 7.92 ResNet122 95.59 92.57 6.64 7.17 CondLaneNet[13] ResNet18 96.01 93.48 3.18 7.28 ResNet34 95.98 93.37 3.20 8.80 ResNet101 96.24 94.54 3.01 8.82 CLRNet[7] ResNet18 95.04 93.97 3.09 7.02 ResNet34 94.73 93.11 2.87 7.92 ResNet101 97.27 96.33 1.86 3.63 CGANet (本研究方法) ResNet18 96.73 95.24 1.84 4.80 ResNet34 96.02 93.78 1.97 6.14 ResNet101 97.45 96.67 2.76 2.31