Automated data-collection for personalized facial expression recognition in human-robot interaction

Face recognition systems, which attempt to identify the emotions that a person is feeling, have been around for quite some time. Facial expression recognition is the technique of detecting facial expressions based on interpretations of patterns in a picture. Because every person's face is unique, when we apply these methods to pictures of people, we are able to identify their facial expressions as being unique. In this research, we build a web-based data collecting application that is completely automated and includes a virtual avatar to guide users through the procedure. The input data we dealt with included written input in the form of six emotions (anger, disgust, fear, happiness, surprise, and sorrow) plus neutral, as well as video footage with a length of 20 seconds for each. With the use of the data, a customized face expression recognition method based on deep learning architecture known as MobileNets would be developed.