식사보조 로봇 사용자의 시선 방향 추정 딥러닝 모델 기반 식사 의도 인식 시스템

Abstract

Care robots that provide dietary supplements to the elderly and the disabled should be able to intelligently interact with the users’ intention to eat in real-time and provide the food they want. To this end, based on deep learning, we have developed a user interface that facilitates the selection of food to be eaten according to six user gaze directions. A single-stage object-detection deep learning model has been implemented as a lightweight, single-stage object-detection deep learning model that can be installed on commercial tablets, and the accuracy of this model was 0.9857.