Fig. 1. Framework of the method

Fig. 1. Framework of the method

Fig. 2. The parameters of the network and illustration of LaneConv and LaneDeconv

Fig. 3. Depth separable convolution. (a) Channel by channel convolution; (b) Pointwise convolution

Fig. 4. (a) Laneconv structure; (b) Lanedeconv structure

Fig. 5. (a) Channel attention; (b) Spatial attention

Fig. 6. DBSCAN cluster

Fig. 7. The output in different stages. (a) Binary output; (b) Clustering output; (c) Fitting output

Fig. 8. Comparison between visualization results of baseline and our method on TuSimple. (a) Original scene; (b) True value; (c) Lanenet results; (d) Results of our method

Fig. 9. Comparison of effects before and after adding CBAM. (a) Not joined CBAM; (b) Joined CBAM

Fig. 10. Visual results generated by our method on some of typical scenarios