In this project, I applied the mehtod proposed in http://dx.doi.org/10.1016/j.acha.2015.11.005 to predict the identity of 6 disputed paintings (maybe Raphael's) and reach a accuary of about 96.5% on leave-one-out cross validation, using tight frame feature extraction and combining Distance Discriminant Analysis and Neural Network.
The functions or usage of the py files can be found in 'usage.txt'