Not to start a flame war - but I wonder if this is one those cases where their vision based system failed to reconstruct the 3d scene. And if that's true, would

Not to start a flame war - but I wonder if this is one those cases where their vision based system failed to reconstruct the 3d scene. And if that's true, wouldn't having Lider helped here? In the long run I agree that vision based system can perform just as good, but I think the deep neural networks Tesla are using is not there yet, and could use supplementary sensors such as Lidar data to improve the training set of the model.