DEEPFAKE Image Synthesis for Data Augmentation

Deepfake videos and images have increased in recent years at exponential rates, and refer to the transfer of features of interest from source image (or video) to target image (or video), such that the target modality appears to animate the source almost close to the real. The forensic professionals, policy makers, and the public alike are interested in the role of Deepfakes is playing to confuse viewers in spreading misinformation by undermining the truth. In the past decade, computer vision has made significant advances using the latest state-of-art-methods of image processing techniques. Supervised Deep learning models produce super-human results in a variety of computer vision and machine learning applications. The field of medical imaging is scarce in terms of a reliable data set that is extensive enough to train distinct models. One way to tackle this problem is to use a Generative Adversarial Network to synthesise DEEPFAKE images to augment the data. DEEPFAKE images can be a useful means in various applications like translating to different useful and sometimes malicious modalities, unbalanced datasets or increasing the amount of datasets. In this paper the data scarcity has been addressed by using Progressive Growing Generative Adversarial Networks (PGGAN). However, PGGAN consists of convolution layer that suffers from the training-related issues. PGGAN requires a large number of convolution layers in order to obtain high-resolution image training, which makes training a difficult task. In this work, a subjective self-attention layer has been added with convolution layers for efficient feature learning and the use of spectral normalization for training stabilization - the two tasks resulting into what is referred to as Enhanced-GAN. The Enhanced-GAN performance is compared to PGGAN performance using the parameters ofAMScore and Mode Score. In addition, the strength of Enhanced-GAN and PGGAN synthesis is evaluated using the U-net model for segmentation tasks. Dice Coefficient metrics show that U-net trained on Enhanced-GAN DEEPFAKE data optimized with real data performs better than PGGAN DEEPFAKE data with real data.