DeepFake Detection Based VGG-16 Model
Wurood A. Jbara, Jamila H. Soud · 2024
This DeepFake technology allows for the creation of convincing fake images and videos that are indistinguishable from real ones, which is causing widespread concern. This work introduces the state-of-the-art technology for detecting DeepFake videos, while based on the VGG-16 Convolutional Neural Network, ensuring accurate and reliable results. Data augmentation is used for performance improving and reduce the computational resources, also using Multi-task Cascaded Convolutional Networks technique for face detection. With this innovative solution, you can easily identify and detect DeepFake videos that have the potential to mislead, deceive, or manipulate viewers. The experimental results using Celeb-DF and DeepFake Face Mask datasets have greatly improved this model for fake video detectors. These datasets pass through several steps that end with the classification stage for fake face detection, the classification report including several metrics: 94.28% for accuracy, 0.1503 for loss, 0.9428 for Precision, 0.9428 for Recall, 0.9859 for AUC and 0.9428 for F1_score in evaluation model performance. The data augmentation process that used to improve the model performance and reduce computational resources. According to the initial results, VGG19 outperforms other analyzed models with a highest accuracy of 94.28%.