Javascript must be enabled for the correct page display

Learning to grasp objects in highly cluttered environments using Deep Convolutional Neural Networks

Oude Vrielink, Jeroen (2021) Learning to grasp objects in highly cluttered environments using Deep Convolutional Neural Networks. Bachelor's Thesis, Artificial Intelligence.

[img]
Preview
Text
bAI_2021_OudeVrielinkJ.pdf

Download (717kB) | Preview
[img] Text
toestemming.pdf
Restricted to Registered users only

Download (124kB)

Abstract

This paper evaluates the performance of the Generative Residual Convolutional Neural Network (GR-ConvNet), proposed by Kumra et al. (2020), as a grasp inference method. The performance is evaluated by testing the network’s predictive accuracy on a subset of the Cornell grasping dataset, and by conducting several rounds of the isolated, pack, and pile grasping experiments, as proposed by Kasaei et al. (2021), using a UR5e robotic arm. For this purpose, a simulation was developed in PyBullet (Coumans et al., 2020). The GR-ConvNet achieved an accuracy of 97.7% on the Cornell grasping dataset. A grasping success rate of 91.9%, 73.2%, and 55.2% was achieved, and a target success rate of 91.9%, 73.2%, and 55.2% was achieved for the isolated, pack, and pile experiments respectively.

Item Type: Thesis (Bachelor's Thesis)
Supervisor name: Mohades Kasaei, S.H.
Degree programme: Artificial Intelligence
Thesis type: Bachelor's Thesis
Language: English
Date Deposited: 20 Jul 2021 13:21
Last Modified: 20 Jul 2021 13:21
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/25369

Actions (login required)

View Item View Item