Abstract
Grasp detection based on convolutional neural network has gained some achievements. However, overfitting of multilayer convolutional neural network still exists and leads to poor detection precision. To acquire high detection accuracy, a single target grasp detection network that generalizes the fitting of angle and position, based on the convolution neural network, is put forward here. The proposed network regards the image as input and grasping parameters including angle and position as output, with the detection manner of end-to-end. Particularly, preprocessing dataset is to achieve the full coverage to input of model and transfer learning is to avoid overfitting of network. Importantly, a series of experimental results indicate that, for single object grasping, our network has good detection results and high accuracy, which proves that the proposed network has strong generalization in direction and category.
Highlights
Over recent years, deep learning has gained huge breakthroughs in computer vision [1, 2]
YOLOv2 has obvious advantages in detection accuracy and real time. us, we introduce it into the research of grasp detection and utilize its “end-to-end” detection manner to establish a grasp detection network model with the proposed 5 grasp parameters as output
Rough analysis of established network model, it can be known that acquired good detection results lie in two reasons. e first one is that our model adopts directly calculation of the loss and carry out global boundary regression on image to acquire the appropriate grasping rectangle. e other is that our model randomly selects a label value for each image during model training, which means that, after multiple training of dataset, the model predicts an average value for each object. us, for single object grasping, the predicted average value still has a good detection effect
Summary
Deep learning has gained huge breakthroughs in computer vision [1, 2]. Unlike traditional hand-engineered features, deep learning can autonomic learning features from images, to acquire highly abstract and robust visual features via making use of image information to the most extent. As one of the most representative deep learning models, convolutional neural network has become a research hotspot in computer vision, with easy training, high performance, few parameters, and strong generalization. Researchers have attempted to introduce it into research on robotic grasp detection, since its remarkable achievements in target detection [3,4,5,6,7,8,9,10,11,12]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
