Using FER+ Dataset to Implement Deep Capsule Network for Facial Emotion Recognition

The researchers have been directed to use the fresh dataset known as the Facial Emotion Recognition Plus (FER+) Dataset. This has enhanced the efficacy of conventional facial recognition technologies, particularly in analyzing human facial expressions. Consequently, it is essential to improve the existing generation of facial recognition systems to accommodate diverse datasets. The current generation of facial recognition technology predominantly depends on deep learning, which presents challenges due to the inadequacy of existing facial databases. The current study proposes the implementation of the Deep Capsule Network and Light Gradient Boosting Machine algorithms utilizing FER+ to develop a system for recognizing facial emotions. This facial database is accessible at no cost for scholarly purposes and was employed to categorize the available images showcasing a range of facial emotions, aligning with one of the seven fundamental human facial expressions, thereby facilitating the development of more sophisticated facial technology. Following our research findings, the proposed facial emotion detection achieves an accuracy of up to 88.92% owing to its meticulously curated facial database.