Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks

Ashraf Zia; Bernard Tiddeman; Patricia Shaw

doi:10.1007/978-3-319-96728-8_6

Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks

Ashraf Zia, Bernard Tiddeman, Patricia Shaw

Department of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference Proceeding (Non-Journal item)

1 Citation (Scopus)

Abstract

Identification of suitable grasping pattern for numerous objects is a challenging computer vision task. It plays a vital role in robotics where a robotic hand is used to grasp different objects. Most of the work done in the area is based on 3D robotic grippers. An ample amount of work could also be found on humanoid robotic hands. However, there is negligible work on estimating grasping patterns from 2D images of various objects. In this paper, we propose a novel method to learn grasping patterns from images and data recorded from a dataglove, provided by the TUB Dataset. Our network retrains, a pre-trained deep Convolutional Neural Network (CNN) known as AlexNet, to learn deep features from images that correspond to human grasps. The results show that there are some interesting grasping patterns which are learned. In addition, we use two methods, Support Vector Machines (SVM) and hotelling’s T2 test to demonstrate that the dataset does include distinctive grasps for different objects. The results show promising grasping patterns that resembles actual human grasps

Original language	English
Title of host publication	Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings
Subtitle of host publication	TAROS 2018
Editors	Maria Elena Giannaccini, Manuel Giuliani, Tareq Assaf
Publisher	Springer Nature
Pages	64-75
Number of pages	12
ISBN (Print)	9783319967271
DOIs	https://doi.org/10.1007/978-3-319-96728-8_6
Publication status	Published - 26 Aug 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10965 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Keywords

Convolutional Neural Networks (CNN)
Deep learning
Grasping patterns
Transfer learning

Access to Document

10.1007/978-3-319-96728-8_6

Permanent link

Persistent link

Cite this

Zia, A., Tiddeman, B., & Shaw, P. (2018). Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks. In M. E. Giannaccini, M. Giuliani, & T. Assaf (Eds.), Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings: TAROS 2018 (pp. 64-75). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10965 LNAI). Springer Nature. https://doi.org/10.1007/978-3-319-96728-8_6

Zia, Ashraf ; Tiddeman, Bernard ; Shaw, Patricia. / Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks. Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings: TAROS 2018. editor / Maria Elena Giannaccini ; Manuel Giuliani ; Tareq Assaf. Springer Nature, 2018. pp. 64-75 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{1c2f6153ad87441cbf7d5d965dae0b88,

title = "Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks",

abstract = "Identification of suitable grasping pattern for numerous objects is a challenging computer vision task. It plays a vital role in robotics where a robotic hand is used to grasp different objects. Most of the work done in the area is based on 3D robotic grippers. An ample amount of work could also be found on humanoid robotic hands. However, there is negligible work on estimating grasping patterns from 2D images of various objects. In this paper, we propose a novel method to learn grasping patterns from images and data recorded from a dataglove, provided by the TUB Dataset. Our network retrains, a pre-trained deep Convolutional Neural Network (CNN) known as AlexNet, to learn deep features from images that correspond to human grasps. The results show that there are some interesting grasping patterns which are learned. In addition, we use two methods, Support Vector Machines (SVM) and hotelling{\textquoteright}s T2 test to demonstrate that the dataset does include distinctive grasps for different objects. The results show promising grasping patterns that resembles actual human grasps",

keywords = "Convolutional Neural Networks (CNN), Deep learning, Grasping patterns, Transfer learning",

author = "Ashraf Zia and Bernard Tiddeman and Patricia Shaw",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing AG, part of Springer Nature 2018.",

year = "2018",

month = aug,

day = "26",

doi = "10.1007/978-3-319-96728-8_6",

language = "English",

isbn = "9783319967271",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Nature",

pages = "64--75",

editor = "Giannaccini, {Maria Elena} and Manuel Giuliani and Tareq Assaf",

booktitle = "Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings",

address = "Switzerland",

}

Zia, A, Tiddeman, B & Shaw, P 2018, Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks. in ME Giannaccini, M Giuliani & T Assaf (eds), Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings: TAROS 2018. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10965 LNAI, Springer Nature, pp. 64-75. https://doi.org/10.1007/978-3-319-96728-8_6

Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks. / Zia, Ashraf; Tiddeman, Bernard ; Shaw, Patricia.
Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings: TAROS 2018. ed. / Maria Elena Giannaccini; Manuel Giuliani; Tareq Assaf. Springer Nature, 2018. p. 64-75 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10965 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference Proceeding (Non-Journal item)

TY - GEN

T1 - Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks

AU - Zia, Ashraf

AU - Tiddeman, Bernard

AU - Shaw, Patricia

N1 - Publisher Copyright: © Springer International Publishing AG, part of Springer Nature 2018.

PY - 2018/8/26

Y1 - 2018/8/26

N2 - Identification of suitable grasping pattern for numerous objects is a challenging computer vision task. It plays a vital role in robotics where a robotic hand is used to grasp different objects. Most of the work done in the area is based on 3D robotic grippers. An ample amount of work could also be found on humanoid robotic hands. However, there is negligible work on estimating grasping patterns from 2D images of various objects. In this paper, we propose a novel method to learn grasping patterns from images and data recorded from a dataglove, provided by the TUB Dataset. Our network retrains, a pre-trained deep Convolutional Neural Network (CNN) known as AlexNet, to learn deep features from images that correspond to human grasps. The results show that there are some interesting grasping patterns which are learned. In addition, we use two methods, Support Vector Machines (SVM) and hotelling’s T2 test to demonstrate that the dataset does include distinctive grasps for different objects. The results show promising grasping patterns that resembles actual human grasps

AB - Identification of suitable grasping pattern for numerous objects is a challenging computer vision task. It plays a vital role in robotics where a robotic hand is used to grasp different objects. Most of the work done in the area is based on 3D robotic grippers. An ample amount of work could also be found on humanoid robotic hands. However, there is negligible work on estimating grasping patterns from 2D images of various objects. In this paper, we propose a novel method to learn grasping patterns from images and data recorded from a dataglove, provided by the TUB Dataset. Our network retrains, a pre-trained deep Convolutional Neural Network (CNN) known as AlexNet, to learn deep features from images that correspond to human grasps. The results show that there are some interesting grasping patterns which are learned. In addition, we use two methods, Support Vector Machines (SVM) and hotelling’s T2 test to demonstrate that the dataset does include distinctive grasps for different objects. The results show promising grasping patterns that resembles actual human grasps

KW - Convolutional Neural Networks (CNN)

KW - Deep learning

KW - Grasping patterns

KW - Transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85051138073&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-96728-8_6

DO - 10.1007/978-3-319-96728-8_6

M3 - Conference Proceeding (Non-Journal item)

SN - 9783319967271

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 64

EP - 75

BT - Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings

A2 - Giannaccini, Maria Elena

A2 - Giuliani, Manuel

A2 - Assaf, Tareq

PB - Springer Nature

ER -

Zia A, Tiddeman B , Shaw P. Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks. In Giannaccini ME, Giuliani M, Assaf T, editors, Towards Autonomous Robotic Systems - 19th Annual Conference, TAROS 2018, Proceedings: TAROS 2018. Springer Nature. 2018. p. 64-75. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-96728-8_6

Estimating Grasping Patterns from Images Using Finetuned Convolutional Neural Networks

Abstract

Publication series

Keywords

Access to Document

Permanent link

Other files and links

Fingerprint

Cite this