EdgeNet: Dense Instance Segmentation via Contour-Detection and Foreground Clustering
Loading...
Authors
Merz, Nicolas
Date
Type
thesis
Language
eng
Keyword
computer vision , instance segmentation , bin-picking , industrial automation , deep learning
Alternative Title
Abstract
This thesis proposes a novel method for instance segmentation in the Robotic
Bin-Picking problem using a blend of deep learning and classical methods, titled
EdgeNet. Pose estimation is used on sensed data to relay information for usage in
object grasping. Industrial bins tend to be incredibly dense, prompting the interest in
segmentation with the intent of further usage in pose estimation. Using an encoder-
decoder network architecture for estimation of foreground, contours and centroids,
followed by a post-processing algorithm, we have achieved performance that is com-
petitive with Mask R-CNN while both training and evaluating faster. Furthermore,
we have created a benchmark for a new task on the Fraunhofer bin-picking dataset and
opened up further research into exploiting instance segmentation to isolate objects
prior to pose estimation.
Description
Citation
Publisher
License
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.