An Automated Virtual Receptionist for Recognizing Visitors and Assuring Mask Wearing

Abstract

Intelligent virtual agents have many societal uses, specifically in situations in which the presence of real humans would be prohibitive. In particular, virtual receptionists can perform a variety of tasks associated with visitor and employee safety, e.g., during the COVID-19 pandemic. In this poster, we present our prototype of a virtual receptionist that employs computer vision and meta-learning techniques to identify and interact with a visitor in a manner similar to that of a real human receptionist. Specifically we employ a meta-learning-based classifier to learn the visitors' faces from the minimal data collected during a first visit, such that the receptionist can recognize the same visitor during follow-up visits. The system also makes use of deep neural network-based computer vision techniques to recognize whether the visitor is wearing a face mask or not.