Recognizing AI-Generated Faces Using Deep Learning

Artificial intelligence is reshaping the world. This technology is changing the way we handle our daily tasks. Deep learning is one method to go toward artificial intelligence and it so far showed great significance when applied into various areas, from medicine to computer vision. However, deep learning showed also that it can be used to harm or help in fraudulence, depending on how it is applied. One example of this is what is called Generative adversarial learning. In 2016, generative adversarial networks (GANs) were presented to the world, and since then many versions of these networks were developed. A GAN can be defined as machine learning model that consists of two neural networks competing each other. These two networks are called generator and discriminator, and each has different role; one is for generating images from random noise given as input, while the latter is for detecting whether the generated images is real or fake. Finally, this whole process is repeated over huge number of iterations until the generator becomes good enough to generate images similar to those of the dataset.

In 2018, CyleGAN was proposed by NVIDIA, which is a modified version of GAN, in which it consists of two generators and two discriminators that provide of image-to-image translation, i.e., transforming images from domain A to domain B and vice-versa. This network…