Image Classification using CNNs

Purpose: To implement and understand the basics of Convolutional Neural Networks (CNNs) via the task of image classification for hand-written characters. 

Background: While a graduate student at the University of Florida, I completed this project with two other students as a final assignment for a Fundamentals of Machine Learning course (Spring of 2022). The deliverable was a final paper. My team won first place in an extra-credit challenge.  

Source Code: Unavailable due to course restrictions.  

Paper Abstract: There are numerous machine learning algorithms appropriate for classification tasks, each with varying assumptions and complexity. This paper applies multiple models to the problem of handwritten character recognition as a project topic for the Fundamentals of Machine Learning course at the University of Florida. Convolutional neural networks performed the best in initial tests, and were down-selected for continued hyperparameter optimization until achieving a test accuracy greater than 90% for a set of ten characters. A second model was developed to identify filter images that were not in original character set, resulting in a filter accuracy greater than 95%.