A Study of Affective Response Design for Robot Using IAPS and Object Detection Algorithm

Abstract

The International Affective Picture System (IAPS) has been used to study people's emotions when looking at pictures. Using this knowledge, we consider whether the Pepper robot can behave as if it has emotions by designing how it reacts to the object detection results from the camera input. As a preliminary step to make Pepper react emotionally, we examined the possibility of designing affective responses using YOLO (You Only Look Once) to detect objects in IAPS photos. We confirmed that YOLO could detect objects in 61\% of the IAPS images and created a label table that maps YOLO labels to IAPS labels. Based on the results from the label table, we proposed additional processing and training for YOLO. We also tested YOLO's object detection with the Pepper camera input and developed a response that reads the names of the detected objects.