This letter proposes a lightweight DNN model for robotic grasping applications with just 1.5 million architecture parameters. In the proposed model, the location of the target object is estimated using a pairwise template matching method, while the orientation of the object is predicted using depth images and a convolutional neural network (CNN). The feasibility of the proposed model is demonstrated both numerically and experimentally on a NVIDIA Jetson NX developer kit. The experimental results show that the grasping system achieves an accuracy of 96.3% and a running time of 125 ms when on 700 images. Moreover, when applied to practical grasping tasks on 20 unseen objects selected from the Cornell grasping dataset, the system achieves an accuracy of 92.5%, which is comparable to that of existing state-of-the-art methods reported in the literature.

Applying deep neural network models to robot-arm grasping tasks requires the laborious and time-consuming annotation of a large number of representative examples in the training process. Accordingly, this work proposes a two-stage grasping model, in which the first stage employs learning-based template matching (LTM) algorithm for estimating the object position, and a self-rotation learning (SRL) network is then proposed to estimate the rotation angle of the grasping objects in the second stage. The LTM algorithm measures similarity between the feature maps of the search and template images which are extracted by a pre-trained model. While the SRL network performs the automatic rotation and labelling of the input data for training purposes. Therefore, the proposed model does not consume an expensive human-annotation process. The experimental results show that the proposed model obtains 92.6% when testing on 2400 pairs of the template and target images. Moreover, in performing practical grasping tasks on a NVidia Jetson TX2 developer kit, the proposed model achieves a higher accuracy (88.5%) than other grasping approaches on a split of Cornell-grasp dataset.