This repository contains the code to implement Siamese Capsule Networks for face verification.
You will need to follow the following steps:
-
First run python setup.py install. If you have a GPU this should install fine on linux but you may have to install directly on pytorch.org and choose the correct settings for your cuda driver version.
-
Then download "All image with funnelling" from here http://vis-www.cs.umass.edu/lfw/#deepfunnel-anchor and place data/lfw/. Similarly download the AT&T dataset training and test splits from here - https://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html and store in data/att_faces
-
Make sure the names of the folder paths