Face Manipulation Detection in Images using Inception Layers and Graph Convolutional Networks

Recently, there has been an upsurge in the availability and advancement of technologies that can morph images. With the advancement in machine learning and deep learning methods, it is now increasingly difficult to detect which image is real or manipulated generated by the latest technologies. When done nonconsensually, images of manipulated faces can harm people's public image and warrant the development of a method that can contrast authentic images from morphed ones. Therefore, image manipulation detection is paramount for regulating online video frame/image data and protecting their authenticity. This study has generated the latest tampered image dataset and subsequently pioneered an efficient model of a two-fold contribution based on Inception Layers with a Graph Convolutional Network (GCN). The Inception layers consist of several convolutional layers of incrementally increasing kernel sizes that help capture morphological information, which GCN processes to capture the relationship between features, followed by a dense layer module. Further, this study conduct experiments on the proposed dataset and compare the results of the proposed method with existing techniques. The result shows that the proposed approach yields better accuracy than the existing state-of-the-art methods.