This is our final project for Harvard's advanced machine learning course, CS 281. Our project attempts to use persistent homology, a technique used to compute topological features of point clouds, to improve the accuracy of basic convolutional neural networks in image classification tasks.