Final project for Harvard's Deep Learning course. I trained a convolutional neural network (CNN) to detect object outlines within an image, optimized to prevent overfitting, and visualized layer activations.
This project is intended to be run in Google Colab, as it is the most cost-effective platform given the size of the training dataset. At the top of each notebook in src/ is a cell containing the os.chdir function and a base directory variable. Modify the paths to correspond to the Google Drive directory you saved them in. Make sure GPU runtime is enabled for each of the notebooks.