Multiscale Efficient Channel Attention for Fusion Lane Line Segmentation

Figure 1

Test results on KITTI-aug dataset. The rows from top to bottom are input images, ground truth, output from LaneNet, SCNN, ENet-SAD, and ours.