Degree Name
MS (Master of Science)
Program
Computer Science
Date of Award
12-2025
Committee Chair or Co-Chairs
Ahamad Al Doulat
Committee Members
Mohammad Khan, Chelsie Dubay, Jeff Roach
Abstract
Building on the WIRA image dataset, this work introduces the Symbiote Particle, an enhanced dual stream CNN model architecture designed to detect AI-generated patterns when used to train on distinct image categories. Additionally, a methodology is proposed to find a single hyperparameter configuration that helps the Symbiote Particle generalize well not only on a single WIRA image category, but on other additional WIRA image categories. The model-configuration combination achieved an overall mean accuracy of 65%, mean precision of 68%, mean recall of 65%, and mean F1 score of 63% for classifying both real and AI images of all 62 WIRA image categories. 35 of WIRA’s 62 total image categories achieved a mean accuracy of 73%, mean precision of 70%, mean recall of 88%, and a mean F1 of 77%. Finally, three WIRA image categories achieved perfect results with 100% accuracy, precision, recall, and F1. These results establish the Symbiote Particle as an effective solution to detect hyper-realistic AI-generated content.
Document Type
Thesis - embargo
Recommended Citation
McDonald, Andrew I., "What is Real Anymore? A Solution to Detect Hyper Realistic AI-Generated Imagery" (2025). Electronic Theses and Dissertations. Paper 4628. https://dc.etsu.edu/etd/4628
Copyright
Copyright 2025 by Andrew Ian McDonald.
