A Faster R-CNN Approach for Partially Occluded Robot Object Recognition

Many objects in household and industrial environments are commonly found partially occluded. In this paper, we address the problem of recognizing objects for use in partially occluded object recognition. To enable the use of more expensive features and classifiers, a region proposal network (RPN) which shares full-image convolutional feature with detector network is needed. We build our approach based on the recent state-of-the-art Faster R-CNN to increase the recognition capability of partially occluded object. We evaluate our approach on the real-time object recognition and robot grasping. The results demonstrate the effectiveness of our proposed method.