• Nisha R
• Dr K R Shylaja

Creating personalized photo-realistic talking head models or that are able to generate reliable video sequences that represent facial expressions and mimics of an individual. In this research, we provide an idea for creating talking head models with short time taken for training and only a couple of photos (so-called several shot learning). Deep learning techniques have become common place recently resulting in the generation of very realistic fake faces, raising about their possible abuse in different fields. A deep neural network architecture that combines content and trace feature extractors is used in this study to present a unique method for identifying and exposing fraudulent faces. Actually, our system can produce an acceptable result after learning from just one photo (one-shot learning) and adding a few more photos to improves the personalization’s quality. According to this, our design talking heads are deep ConvNets that directly generate video frames through a series of convolutional neural networks rather than by warping them. A technology when dealing with few-shot capabilities is what we give us. It does extensive meta-learning on a huge dataset. The content extractor focuses on extracting high level semantic information and determining patterns that can tell real face characteristics apart from false ones. At the same time, the data extraction explores minute imperfections and discrepancies in facial patterns through the advantage of the trace left behind the generating process.

Deep Learning, Fake Faces, Convolutional Neural Network, Detection

"Disclosing Fake Faces Using Deep Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.10, Issue 11, page no.b43-b51, November-2023, Available :http://www.jetir.org/papers/JETIR2311108.pdf

"Disclosing Fake Faces Using Deep Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.10, Issue 11, page no. ppb43-b51, November-2023, Available at : http://www.jetir.org/papers/JETIR2311108.pdf