Vision based Generous Deep Neural Network Grasping Detector for Robot Grasping Task

Abstract

Rotation invariance has been an important topic in computer vision tasks such as face detection [1], texture classification [2] and character recognition [3], to name a few. The importance of rotation invariant properties for computer vision methods still remains for recent DNN based approaches. In general, DNNs often require a lot more parameters with data augmentation with rotations to yield rotational-invariant outputs. Max pooling helps alleviating this issue, but since it is usually 2 2 [4], it is only for images rotated with very small angles. Recently, there have been some works on rotation-invariant neural network such as rotating weights [5, 6], enlarged receptive field using dialed convolutional neural network (CNN) [7] or a pyramid pooling layer [8], rotation region proposals for recognizing arbitrarily placed texts [9] and polar transform network to extract rotation-invariant features [10]. Applications of deep neural network based object and grasp detections could be expanded, significantly when the network output is processed by a high-level reasoning over relationship of objects. Recently, robotic grasp detection and object detection with reasoning have been investigated using deep neural networks (DNNs). There have been effects to combine these multi-tasks using separate networks so that robots can deal with situations of grasping specific target objects in the cluttered, stacked, complex piles of novel objects from a single RGB-D camera. We propose a single multi-task DNN that yields an accurate detections of objects, grasp position and relationship reasoning among objects. Our proposed methods yield state-of-the-art performance with the accuracy of 98.6%and 74.2% with the computation speed of 33 and 62 frame per second on VMRD and Cornell datasets, respectively. Our methods also yielded 95.3% grasp success rate for novel object grasping tasks with a 4-axis robot arm and 86.7% grasp success rate in cluttered novel objects with a humanoid robot