Using deep learning to detect deepfake videos
Main Article Content
Abstract
In recent times, software based on deep learning, due to their ease of availability have made the modelling of real- looking face swapping in videos very easy that leave little signs of forgery. Such forged videos are termed deepfake(DF) videos. Manipulation of digital videos has been illustrated for many decades via an adequate usage of visual effects. Recent progress in the field of deep learning has caused an abrupt increase in the realism of forged content and they can be made very conveniently. These AI-produced means are also called by DF. Forming the DF with the artificial intelligence tools is an easy task. Developing an application to find out whether the given video is a deepfake isn’t an easy thing to do. This is mainly because any such algorithm’s training will require lots of computations. To accomplish this daunting task, we decided to use convolutional neural networks and recurrent neural networks. We begin by extracting features at the frame level using CNN. We will train a recurrent neural network using those extracted features. Our RNN will then be able to identify if a video is fake or not and also check for any temporal variations between frames which were caused by the deepfake forming tools. We will compare the performance of our model with some results from a standard data set. We will keep on improvising our model till it becomes good enough to work with real world data.
Downloads
Metrics
Article Details
Licensing
TURCOMAT publishes articles under the Creative Commons Attribution 4.0 International License (CC BY 4.0). This licensing allows for any use of the work, provided the original author(s) and source are credited, thereby facilitating the free exchange and use of research for the advancement of knowledge.
Detailed Licensing Terms
Attribution (BY): Users must give appropriate credit, provide a link to the license, and indicate if changes were made. Users may do so in any reasonable manner, but not in any way that suggests the licensor endorses them or their use.
No Additional Restrictions: Users may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.