Real-time object detection-based affordance feature extraction system using deep learning model

Abstract

In this paper, we deal with a method for extracting the features needed to estimate the object's affordance based on the results of the object detection process using a deep learning model. To implement the object detection process, we utilize the You Only Look Once (YOLO) network, which is known to be the fastest and most accurate among the latest deep learning models. To extract the affordance features from the image within the results of object detection, the following steps are sequentially performed: image segmentation, binary conversion, and principal component analysis for the object image within the bounding box of the target object. The simulation results for the proposed method are shown. © ICROS 2017.