Real and Fake Face Detection: A Comprehensive Evaluation of Machine Learning and Deep Learning Techniques for Improved Performance

The progress of digital manipulation methods has led to the creation of incredibly realistic fake faces, making it more challenging for humans to differentiate between genuine and fabricated ones while our brains are wired to interpret facial features, the use of sophisticated technology and artificial intelligence is blurring the line between real and manipulated images. As a consequence, techniques such as deep learning are becoming popular in distinguishing between real and fake faces with greater accuracy and reliability. In this study, a combination of machine learning and deep learning approaches was utilized to develop models for the detection of genuine and fabricated faces. The initial model implemented was an artificial neural network model, which utilized a Fourier-based technique for feature extraction. The model underwent training and testing using a benchmark dataset labeled as "Real and Fake Faces", and produced an accuracy rate of 0.57. ResNet18 approach involved the utilization of multiple models of convolutional neural networks that were trained on the same dataset and their results enhanced the overall precision of the classification process. The implementation of ResNet18 in deep learning has greatly enhanced the overall performance by achieving a substantially elevated accuracy level of 0.77.