Zehan Zheng

The currently organized code only provides training and testing for the OpenLane dataset. We will update the support for the Apollo dataset in the future, thank you for your attention.

We update the model code and corresponding checkpoint, so you need to download the newest file from [here](https://drive.google.com/file/d/1QE5Vt6jJl1uyduEi5tj7_0iN6-_Mpqbp/view?usp=sharing).

Thank you for your issue. In `eval_3D_lane.py` we have done some pre-processing of the gt lanes to keep only the **visible** portion and **within the sampling range** (the processing is...

按照报错应该是传入的图片为空导致的，你可以检查一下摄像头实时传入的图像是不是一直有效。

并不行。除非知道标定板在世界坐标系下的位置才能计算，比如传统的方法用全景标定布来标定相机外参，然后再逆投影变换到BEV视角。

It should be `self.time_seq[len(input)//2 - 1]` or `self.time_seq[len(input)//2]`. We'll revise it later, and thanks for pointing out the mistake.