Face Representation for Online Interactions Using Bidirectional Generative Adversarial Networks (BiGANs)
Bidirectional Generative Adversarial Networks (BiGANs), Face Representation, Efficient Data Transmission, Online Interactions, Deep Learning.
In this research, a new method for face representation is presented, which utilizes Bidirectional Generative Adversarial Networks (BiGANs), showing significant progress compared to conventional methods of video transmission using MPEG-2 compression techniques. In scenarios such as online meetings, our approach takes advantage of the inherent bidirectional capabilities of BiGANs in virtual environments to produce compact yet highly expressive facial representations. As a result, the amount of data required for transmission is reduced. The effectiveness of our approach in generating high-quality synthetic face images that closely resemble the original faces was demonstrated through our experiments, which were conducted on a dataset consisting of 813 frames of an individual's face. Furthermore, the method's capability to preserve higher values of the Structural Similarity Index (SSIM) and Peak Signal-to-Noise Ratio (PSNR) highlights its potential to generate synthetic facial images with minimal degradation in quality. This makes it an encouraging approach for real-time online communication, especially in situations with limited bandwidth.