In this project we primarily focused on employing adequate data augmentation to a somewhat small dataset in order to provide more and harder data points to our models so that they may be less overfitted. We also briefly explored the concept of ensemble learning through hard and soft voting, where we obtained a final result of 99.4% accuracy which is decent, but still leaves some room to improve, especially in using more varied models in our ensemble in order to cover a wider variety of outliers