As part of a graduate-level machine learning course, my group and I decided to classify nanoscale barcode images using a variety of classical and novel machine learning techniques. Using techniques such as Naive Bayes, Random Forest, Extreme Gradient Boosted Random Forest, Support Vector Machines with multiple kernels, and Convolutional Neural Networks (CNN), we were able to classify the images at around 75% accuracy with CNNs. Some of the challenges we faced included extremely noise data (some of the barcodes physically overlapped with each other) and efficiently training our models on an extremely large dataset.

A full project summary will be made available as soon as I am able.