Clarification about question

In order to train the parameters of a face recognition system, it would be reasonable to use a training set comprising 100,000 pictures of 100,000 different persons.

Using a siamese network don’t we just need a single image of a person ?

Can you say more about the context of this question? Was it in one of the quizzes?

Yes, this was in one of the quizzes, it was true/false question and I think it should be true

This is covered in the “one shot learning” video.

The problem stated in the quiz question is that having just one image for each person is not enough to learn a reliable identification.

@krithika_govindaraj in my mind, with the Siamese network we need to establish a ‘degree’ of difference-- But, figure, if all the images in the set are different, then we are extracting a difference from, what, exactly ? You’ve no point for confirmation/comparison.

Perhaps you could say person A is not B, though not A is A, for sure, without more information. (*Would just add if you have 100,000 images of 100,000 people, kind of a priori you know there won’t be a match. That conclusion doesn’t even require any computation.)

Thus the need for a positive and negative reference.