Building a NN that classifies images based on human preference

vision
1

Hi,
I have a data set containing triplets of three images of food A, B and C where it is given that a human thinks A and B are closer in taste than A and C.
I want to build a neural net that learns from these preferences and will be able to predict, if given a triplet of images of food, if A & B or A & C are closer in taste.

Does any one have any ideas on how I might do this?
I am new to machine learning so any suggestions on network architecture or loss function would be most helpful

2

You could look siamese networks and triple loss to start

1 Like