Video input pipeline

[WilliamLockeIV]

Hi! Thanks for the great work on this.

I just wondered, is there a pipeline for taking video (e.g. from a dashcam) and feeding it into the lanenet model for inference? I'm sure we could implement something like that on our own using OpenCV, but I wasn't sure if I'd missed it somewhere in the code already.

You also mention in the ReadMe that while the deep neural net can do inference at 50 fps (78 fps with BiseNetV2), the input pipeline needed to be improved to achieve real time lane detection. Where does the bottleneck seem to be?