The master’s thesis by the student Doaa Alaa Talib, a student at the University of Basra, College of Engineering, Department of Computer Engineering, was discussed. It included research on the method of generating fake images and video clips and the method of detecting them using different algorithms. GAN networks were used to create fake images using an algorithm, StyleGAN2-ADA, which outperformed StyleGAN2. .
The FFHQ dataset, which was trained on an NVIDIA RTX 2070 processor, showed increased accuracy with more data. The system succeeded in dealing with personal images, implementing real-time use, and developing an application for image processing and detection.
The researcher found that StyleGAN2-ADA showed advantages in data efficiency, clarity, and accuracy in detecting fake images, and achieved a success rate of 95.71%.