Lets say for example, i want a none class in MNIST.
classes: 0,1,2,3,4,5,6,7,8,9,None.
How do i choose/make data for such a class(None)? Not programatically , but what kind of data should i make so that classifier can differentiate between (0-9) and None.
Not only MNIST, CIFAR10,…etc.
The easiest way would be to simply add one class label (e.g. 10) and define this label to be equivalent to None.
However, this could not be applied to the datasets you mentioned, because the datasets don’t contain such elements (they are build with their specific classes in mind), so you would have to add these samples by yourself.
Another approach would be to do something like here.