Recent advances in machine learning and deep learning have brought great improvement in our ability to solve complex problems from computer-vision, speech recognition to self-driving cars in our advancement towards true AI. However, these advances in deep learning are also being used to create software that can endanger, cause threats to people and our society. Face image manipulation techniques can create fake images which are very much impossible to detect with the human eyes. Therefore, there is a need to be able to detect these fake images automatically.
This research presents how to detect fake images by making use of transfer learning with a generalized pre-trained convnet to detect fake images. Some of the most popular fake image manipulation techniques and attempts are looked at in this research to detect and classify them into their correct classes.
Click here to read the full document.
View the code here