Please use this identifier to cite or link to this item: http://dspace.dtu.ac.in:8080/jspui/handle/repository/20798
Title: DESIGN AND DEVELOPMENT OF FRAMEWORK FOR DEEPFAKE VIDEO DETECTION USING CNN AND LSTM
Authors: DAGAR, NIKITA
Keywords: DEEPFAKE VIDEO DETECTION
GENERATIVE ADVERSARIAL NETWORKS (GAN)
FRAMEWORK
LSTM
CNN
Issue Date: May-2024
Series/Report no.: TD-7316;
Abstract: The increasing worldliness of mobile and camera based techniques which have been seen to have expanding scope of multi-media content and the internet based facilities have made it more convenient than ever before to produce and distribute digital videos. Although the manipulations of digital images andvideos have been seen by the use of visual effects for several decades, recent developments in deep learning have caused a drastic increase in the real content and accessibility of the manipulated content that can be produced. Recent advancements in automated video and audio editing tools, Generative Adversarial Networks (GAN) and a good and manipulated video content obtained from internet or social media can be produced and easily disseminated. In todays scenario, it has become difficult for people to distinguish the fact from fiction because of the technological advancements made in automated approaches for creating and sharing content online. A survey of videos, often indecent, the face of one person in the source video is swapped with face of other person in the resultant video using deep neural network that automatically maps the face expressions of the person in the original video to the expression of other person in manipulated video, so called deepfakes which are grabbing a lot of widespread alarm. Deepfakes are named so because they use artificial neural network and deep learning to create fake content. The deep learning based models have been widely used for creating manipulated content and some of the models are Autoencoders and Generative Adversarial Networks (GAN). These models monitors face expressions and different kinds of movement of the person in the original video and then synthesize face images of the another person in the manipulated video to make similar face expressions and other movements. Nowadays creating deepfake is quite easier because of the developments of applications like faceapp and fakeapp, nowdays anyone can use these application in order to create their own manipulated content. So, it has become very important to detect deepfakes in order to avoid the spread of the fake content. This report presents the survey of algorithms used to create deepfake and deepfake video content detection methods.
URI: http://dspace.dtu.ac.in:8080/jspui/handle/repository/20798
Appears in Collections:M.E./M.Tech. Information Technology

Files in This Item:
File Description SizeFormat 
NIKITA DAGAR M.Tech.pdf2.34 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.