UVA Building Classifier

Building an Image Classifier to Correctly Identify UVA Buildings


This project was completed for Dr. Stephen Baek's Deep Learning Class as part of the UVA MSDS residentia program. I implemented a range of convolutional neural networks with the ultimate goal of correctly classifying images of UVA buildings into one of 18 categories. I created and trained two models from scratch, both inspired by the VGG architecture, and then capitalized on transfer learning of a RESNET architecture. While all three models had test accuracies of between 80 and 90%, my final model, an ensemble of all three, achieved >99% accuracy on the training set and >95% accuracy on the test set.