Implementation of our paper - Common Representation Learning Using Step-based Correlation Multi-Modal CNN presented as spot-light paper in Asian Conference on Pattern Recognition (ACPR-2017).
This file contains the code for CorrMCNN (2nd architecture). The 7th loss variables can be removed to obtain the 1st architecture. Furthermore, dropout and batch normalization layers have been removed in the above version of code, which can be added by the viewer.