My knee-jerk is to be upset because the risk factor of a poorly functioning pedestrian classifier is obviously higher than something like a sentiment analyzer, but at the same time, this dataset is just for educational purposes, right? Is Udacity actively recommending people use this in production settings?