News for hackers. Stuff that matters.
Re:Slight Street Sign Modifications Can Fool Machine Learnin... by geophile on Saturday August 05, 2017 @10:47AM (#14935973)
This article says that the training set was small. And presumably the misclassified images were unlike images in the training set. If the training set includes images intended to mislead, then wouldn't the classifier then be more tolerant?

There are no comments on this comment yet.