Abstract: The project “Deepfake Detection Using Xception and Mobilenets Deep Learning Models” is a web-based application for identifying deepfake media contents i.e., image and video using deep learning technologies. Deepfake can be simply defined as “an image or video of a person in which their face or body has been digitally altered so that they appear to be someone else”. It is a controversial technology with many wide-reaching issues impacting society, e.g., election biasing. The existing system is based on cross-domain fusion, which works on the basis of traditional spatial domain features. This method had utilized the publicly deepfake datasets, and the results show that the method is effective particularly on the Meso-4 Deepfake Database. But this system is only capable of analysing the spatial features, so we propose a system that can process both image and video input and performs both spatial and depth-wise analysis over the input data. The deep learning models Xception and Mobile Net are the two approaches used for classification tasks to detect deepfakes. We utilize training and evaluation datasets from Face Forensics++ comprising four datasets, Face swap, Face2Face, Deepfake, Neural Texture generated using four different and popular deepfake technologies. The input is analysed for both spatial and depth features which is made possible through Xception and Mobile nets that uses depth wise convolutions. It is capable of detecting almost all kind of deepfakes since we train our model with dataset that contains the data obtained from popular deepfake creation.
Keywords: Deep learning, Web, Database, Texture.
Works Cited:
D. Rupasri, M. Kumaran, J. Lin Eby Chandra " Deepfake Detection Using Xception and Mobilenets Deep Learning Mod", IJARCCE International Journal of Advanced Research in Computer and Communication Engineering, vol. 12, no. 9, pp. 95-99, 2023. Crossref https://doi.org/10.17148/IJARCCE.2023.12916
|
DOI:
10.17148/IJARCCE.2023.12916
