Deepfake Detection Using Xception and LSTM
International Research Journal of Modernization in Engineering Technology and Science · 2023
Deepfakes are fabricated works of art where a person appearing in an earlier photograph or motion picture is modified to exhibit the face of another person.A deep learning-based generative model called a Generative Adversarial Network (GAN) is a model architecture for training generative models to create fake videos.In the context of GANs, the generation model lends significance to points in a predetermined latent space, enabling fresh points pulled from the latent space to be fed to the generator model as input and utilized to produce brand-new and distinctive output instances.Due to this, it is simple to use GANs to build deep fakes that can be used inappropriately in various contexts.Due to the engrossing misuse of deepfakes, there is a need for appropriate detection tools.Deepfake detection requires a large amount of data to train models and test them.They require large datasets that contain thousands of videos, both real and fake at equal ratios to avoid biased results.This paper works with deep learning-based deep fake detection using Convolutional Neural Network(CNN) and Recurrent Neural Network(RNN), CNN utilizing the Xception network, and ytLSTM as RNN.In the algorithm, the spatial features are recognized by the Xception and LSTM identifies the temporal inconsistency between frames.The models are trained against three standard datasets.The results are also validated with standard datasets resulting in deepfake prediction with a minimal computational time and nominal accuracy.