Single-grasp detection based on rotational region CNN

Shan Jiang, Xi Zhao, Zhenhua Cai, Kui Xiang, Zhaojie Ju

Research output: Chapter in Book/Report/Conference proceedingConference contribution

168 Downloads (Pure)

Abstract

Object grasp detection is foundational to intelligent robotic manipulation. Different from typical object detection tasks, grasp detection tasks need to tackle the orientation of the graspable region in addition to localizing the region since the ground truth box of the grasp detection is arbitrary-oriented in the grasp datasets. This paper presents a novel method for single-grasp detection based on rotational region CNN (R2CNN). This method applies a common Region Proposal Network (RPN) to predict inclined graspable region, including location, scale, orientation, and grasp/non-grasp score. The idea is to deal with the grasp detection as a multi-task problem that involves multiple predictions, including predict grasp/non-grasp score, the inclined box and its corresponding axis-align bounding box. The inclined non-maximum suppression (NMS) method is used to compute the final predicted grasp rectangle. Experimental results indicate that the presented method can achieve accuracies of 94.6% (image-wise splitting) and 95.6% (object-wise splitting) on the Cornel Grasp Dataset, respectively. This method outperforms state-of-the-art grasp detection models that only use color images.
Original languageEnglish
Title of host publicationAdvances in Computational Intelligence Systems
Subtitle of host publicationContributions Presented at the 19th UK Workshop on Computational Intelligence, September 4-6, 2019, Portsmouth, UK
EditorsZhaojie Ju, Longzhi Yang, Chenguang Yang, Alexander Gegov, Dalin Zhou
PublisherSpringer
Pages131-141
ISBN (Electronic)978-3-030-29933-0
ISBN (Print)978-3-030-29932-3
DOIs
Publication statusPublished - Sep 2019
Event19th UK Workshop on Computational Intelligence - Portsmouth, United Kingdom
Duration: 4 Sep 20195 Sep 2019
Conference number: 19
https://www.ukci2019.port.ac.uk/

Publication series

NameAdvances in Computational Intelligence Systems
PublisherSpringer
Volume1043
ISSN (Print)2194-5357
ISSN (Electronic)2194-5365

Workshop

Workshop19th UK Workshop on Computational Intelligence
Abbreviated titleUKCI 2019
Country/TerritoryUnited Kingdom
CityPortsmouth
Period4/09/195/09/19
OtherThe UKCI 2019 covers both theory and applications in computational intelligence. The topics of interest include
Fuzzy Systems
Neural Networks
Evolutionary Computation
Evolving Systems
Machine Learning
Data Mining
Cognitive Computing
Intelligent Robotics
Hybrid Methods
Deep Learning
Applications of Computational Intelligence
Internet address

Fingerprint

Dive into the research topics of 'Single-grasp detection based on rotational region CNN'. Together they form a unique fingerprint.

Cite this