Andrew Ng mentioned that we need 2N+1 positions for Landmark detetcion because we always need a variable (face?) predicting if this is a face first, then positions for l1x, l1y…lnx,lny, right?
TMosh
April 10, 2023, 7:09pm
2
Please provide the video name and time mark where Andrew discusses this.
TMosh
April 11, 2023, 3:32pm
3
This was discussed in a previous thread (found by searching for the keyword “landmark”):
Hi,
One of the questions in the C4 W4 Quiz asks about the number of output units of the landmarks.
"Question:
When building a neural network that inputs a picture of a person’s face and outputs N landmarks on the face (assume that the input image contains exactly one face), we need two coordinates for each landmark, thus we need 2N output units. True/False?"
The answer is marked as True. However, doesn’t the output need one more unit telling the probability of whether there is a face or not?…
