Download the full-sized PDF
Permanent link (DOI): https://doi.org/10.7939/R3X05XP9M
This file is in the following communities:
|Faculty of Graduate Studies and Research|
This file is in the following collections:
|Theses and Dissertations|
Learning Deep Representations, Embeddings and Codes from the Pixel Level of Natural and Medical Images Open Access
- Other title
- Type of item
- Degree grantor
University of Alberta
- Author or creator
Kiros, Ryan J
- Supervisor and department
Csaba Szepesvari (Computing Science)
- Examining committee member and department
Linglong Kong (Statistics)
Russell Greiner (Computing Science)
Department of Computing Science
Statistical Machine Learning
- Date accepted
- Graduation date
Master of Science
- Degree level
Significant research has gone into engineering representations that can identify high-level semantic structure in images, such as objects, people, events and scenes. Recently there has been a shift towards learning representations of images either on top of dense features or directly from the pixel level. These features are often learned in hierarchies using large amounts of unlabeled data with the goal of removing the need for hand-crafted representations.
In this thesis we consider the task of learning two specific types of image representations from standard size RGB images: a semi-supervised dense low-dimensional embedding and an unsupervised sparse binary code. We introduce a new algorithm called the deep matching pursuit network (DMP) that efficiently learns features layer-by-layer from the pixel level without the need for backpropagation fine tuning. The DMP network can be seen as a generalization of the single layer networks of Coates et. al. to multiple layers and larger images. We apply our features to several tasks including object detection, scene and event recognition, image auto-annotation and retrieval. For auto-annotation, we achieve competitive performance against methods that use 15 distinct hand-crafted features. We also apply our features for handwritten digit recognition on MNIST, achieving the best reported error when no distortions are used for training. When our features are combined with t-SNE, we obtain highly discriminative two dimensional image visualizations. Finally, we introduce the multi-scale DMP network for domain independent multimodal segmentation of medical images. We obtain the top performance on the MICCAI lung vessel segmentation (VESSEL12) competition and competitive performance on the MICCAI multimodal brain tumor segmentation (BRATS2012) challenge.
We conclude by discussing how the deep matching pursuit network can be applied to other modalities such as RGB-D images and spectrograms.
- Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.
- Citation for previous publication
Kiros, R., & Szepesvari, C. (2012). Deep Representations and Codes for Image Auto-Annotation. In Advances in Neural Information Processing Systems 25 (pp. 917-925).Kiros, R., & Szepesvari, C. (2012). On Linear Embeddings and Unsupervised Feature Learning. International Conference on Machine Learning, Representation Learning Workshop.
- Date Uploaded
- Date Modified
- Audit Status
- Audits have not yet been run on this file.
File format: pdf (Portable Document Format)
Mime type: application/pdf
File size: 5852178
Last modified: 2015:10:12 19:25:19-06:00
Original checksum: 7e620fb4692fa927307b3ff794f310da
Well formed: false
Status message: Lexical error offset=5843221
Page count: 89