This project is made by me as a part of Seasons of Code (SoC) program of IIT Bombay.
In this project I used Principal Component Analysis (PCA) to reduce the dimensionality of training set of face images which made it very efficient for comparing the test images with each category.
In the basic implementation, face_recognition.m, I simply calculated the mean squared distance and assigned it to the category from which the mean distance is minimum. This approach gave an accuracy of about 71%.
In a more complex implementation, face_recognition_nn.m, after reducing the dimensions, I trained the reduce images in a neural network and then passed the tets images through the trained neural network from classification. This increased the accuracy to about 98%.