[Python Glasgow] Training a classifier on pairs of images rather than one image

[Abdul Abdul]

Hello,

I hope you are doing fine, and apologize for my disturbance. I just have a
question and thought you might have an idea on it. I was walking through this
example
<https://www.pyimagesearch.com/2017/12/11/image-classification-with-keras-and-deep-learning/>,
which is clear enough.

My question is how can I train the network on "pairs" of images rather than
single images.

Say for instance that I have two classes: "similar" and "nonsimilar". Say
that in the similar class I have a group of subdirectories, such that each
subdirectory contains two images considered similar, and the same for the
"nonsimilar" class, but each subdirectory will contain two images that are
not considered similar.

I'm not able to figure out how to train the example in the link above on
the scenario mentioned in the above paragraph, and thought you might have
an idea.

Thanks so much for your kind support.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/glasgow/attachments/20180113/a7dacfcd/attachment.html>