Problem when do inference on CurveLanes

[han9608]

I downloaded curvelanes.pth and config, running demo.py, but the result does not meet my expectation.

Is there anyone know the question?(curvelanes data process is different from others?)

[han9608]

已解决（偷懒方法，临时使用）：

pred2coord 函数传一个cfg,dataset，下面改成这个样子：

if dataset == 'Tusimple' or dataset == 'CULane':
    row_lane_idx = [1, 2]
    col_lane_idx = [0, 3]
else:
    row_lane_idx = [3, 4, 5, 6]
    col_lane_idx = [2, 7]

变量含义不明确，有知道的大佬请指导一下（为什么车道只检测出两条）