Hi, I have a question. I try to classification object with CNN. I read that the output(feature vector) of the model (sometimes) should be equal to the number of classes, but if I want the output to be a feature vector of 256(size) and the number of classes is 500. How compare the output with the target?
Usually you are using an output layer with the same number of outputs as your number of classes.
If you need a feature vector in another shape, you could use the e.g. penultimate layer output for it (sometimes called FC7 features IIRC).
Thank you for your answer 