GNF: Generative Natural Faces Dataset for Face Expression Recognition Task

In computer vision, Facial Expression Recognition (FER) is a critical task in various applications, such as human-computer interaction. The effectiveness of FER systems depends mainly on the quality and diversity of the datasets used. The traditional FER data sets may not fully satisfy these requirements due to the lack of diversity, class imbalance, and mislabeling. To address some of these issues, we introduce a novel facial expression dataset created by using Generative Adversarial Networks (GANs). To evaluate its efficacy, the training and test capabilities of the dataset on various CNN architectures are compared with benchmark FER datasets.