Let’s say I have a label map with 0: background and 1: label
I am implementing an autoencoder that takes the input label and extends it to be a multi-label where each label is converted into a separate channel.
Now my question: When using the one hot encoder
with a label map that has only one label (1), would the number of classes that I put into the on hot encoder
be C=1 or C=2?
I am not sure if the one hot encoder
converts the 0 values in its own channel. Would it?
Thanks a lot for your response.