In the video on Face recognition and binary classification when Andrew talks about pre-computing the vectors for the images and then just passing it to the network to predict if it is the same person I am confused as to how the network will know which of the pre-computed vectors it will use to compare with a fresh image. I’m assuming the pre-computed vectors would be stored in memory somewhere and the new image would be compared to each of these right?
Or have i got something wrong?
Could someone clarify please.
Yes, that is how it works.