Facial Forgery Detection

Abstract

In the recent past, fields of computer vision and graphics have become more advanced and it is now easy to generate realistic fake videos and images. At present facial forgery techniques, such as Deepfake and Face2Face are very popular. Such forged videos and images can be used for fake news and to deceive biometric recognition systems. Detection of facial forgery in images and videos is a very complex and challenging task. For the last years, different methods have been employed to detect such forgeries. However such facial forgeries are still challenging to detect. In this paper, we proposed a facial forgery detection solution which automatically detects facial forgeries in videos and images. The proposed technique will classify fake and original videos and images. To classify the images we used Inception-ResNet, a deep neural network. The proposed technique is validated on publicly available Deepfake TIMIT dataset and reported very effective results. i