Deep learning for picking point detection in dense cluster

This paper considers the problem of picking objects in cluster. This requires the robot to reliably detect the picking point for the known or unseen objects under the environment with occlusion, disorder and a variety of objects. We present a novel pipeline to detect picking point based on deep convolutional neural network (CNN). A two-dimensional picking configuration is proposed, thus an extensive data augmentation strategy is enabled and a labeled dataset is established quickly and easily. At last, we demonstrate the implementation of our method on a real robot and show that our method can accurately detect picking point of unseen objects and achieve a pick success of 91% in cluster bin-picking scenario.