DeepFake Videos Detection Using Machine Learning

With the advancement of technology and machine learning the feasibility for people to produce fake videos and spread fake news is becoming more and more easy especially with the creation of GAN models which are capable of producing accurate enough fake videos. The main agenda of this project is to help detect these fake videos from real using machine learning techniques. In this project the data from faceforencis was considered and frames are generated from videos followed by application of novel technique in classification by using face detection over videos. These images are transformed using various parameters and models are compared against them. Highest accuracy of 95% with AUC value of 0.95 is obtained using laplace transformed images. Quantum machine learning is also applied and was able to train the model with 50% less time than classical model