Degree
Master of Science in Data Science
Department
Department of Computer Science
Faculty/ School
School of Mathematics and Computer Science (SMCS)
Date of Submission
Fall 2025
Supervisor
Dr. Muhammad Atif Tahir, Professor and Program Coordinator, Graduate & Postgraduate Programs (CS)
Co-Supervisor
Dr. Rizwan Ahmed Khan
Keywords
ResNet, Synthetic, AI, cross-generator, augmentation, artifacts, transferability
Abstract
This project “Real Versus Synthetic Image Classification”, focused on the use of deep convolutional networks for differentiating AI generated images from real photographs. We explored ResNet based architectures to explore the effectiveness of residual learning for robust feature extraction and generalization. Our experiments investigate two complementary training streams designed to evaluate how dataset diversity and model capacity impacts detection reliability. Through systematic training, augmentation, and inference strategies, we examine the ability of ResNet models to identify subtle artifacts introduced by generative models. The results highlight that deeper residual architectures combined with diverse training data significantly improves detection accuracy and robustness against different unseen AI sources. We conclude with insights on improving transferability and cross-generator generalization in real–synthetic image classification.
Document Type
Restricted Access
Submission Type
Research Project
Recommended Citation
Raza, H. (2025). Real vs. Synthetic Image Classification Using Deep Learning Do you like this personality? (Unpublished graduate research project). Institute of Business Administration, Pakistan. Retrieved from https://ir.iba.edu.pk/research-projects-msds/67
The full text of this document is only accessible to authorized users.
