Fake Image Detection

This project aimed to detect fake face images, such as deep fakes and photoshopped images, from real ones. Different CNN architectures, and various image processing transforms, such as the Fourier transform, were used to tackle this problem. Furthermore, the responses from CNN layers were utilized to localize the edited region. The final achieved performance was around 0.94 f1-score on the test sets.

• Tech Stack: Python, TensorFlow, OpenCV, Keras