Graduate Thesis Or Dissertation
 

Flex-shape convolution and an interactive image segmentation system

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/5d86p434r

Descriptions

Attribute NameValues
Creator
Abstract
  • Semantic image segmentation is a relatively difficult task in computer vision. With the advent of deep learning, semantic image segmentation is increasingly of interest for researchers because of the excellent predictions from Convolutional Neural Network (CNN). However, CNNs have proven to struggle with obtaining global context of image due to convolutions being a local operation. Recent researches have proposed several global context-aware approaches: Atrous convolution, V-Net, multi-scale architecture, etc. Different from previous perspectives, this paper proposes two novel approaches to solve the problem of locating context in CNN. The first approach is called flex-shape convolution that improves the scaling and deformation capabilities of the CNN. The second approach samples auxiliary positive object pixels and negative non-object pixels for network to infer the object mask. Additionally, a web-based application of interactive image semantic annotator has been developed that allows both classical image segmentation algorithms and deep learning algorithms to assist researchers with annotating their images for semantic segmentation.
  • Keywords: Data science, Interactive image segmentation, Deep learning, Software development
License
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Non-Academic Affiliation
Rights Statement
Publisher
Peer Reviewed
Language
Replaces

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details ZhouZheng2017.pdf 2017-10-20 Public Download