We propose an unsupervised task to achieve better generalisation. The task would be to focus on salient features in the image and help our models learn between what to focus on in data and what not to. This has been achieved by training on more samples in an unsupervised way.
With the advent of deep learning the question of what our models are learning is of utmost importance. Is it learning to mimic the data seen during training or it is learning important features from the data is a question we seek to answer. For example if we take a grayscale image of a flower our human eye would be able to still classify it correctly. It’s said that we as humans learn and learn to generalize in an unsupervised fashion growing up.In the course of this project we will analyse if this could be true for deep learning models as well.
MSE(I' and I)