Image Similarity Grid

Jan Hartman (2018) Image Similarity Grid. EngD thesis.

Preview

Abstract

The thesis describes the development of an interactive image visualization tool. To improve the display of images, the tool combines the use of deep learning with other data analytics techniques. The proposed solution consists of several sequentially connected methods. The first step is using convolutional neural networks for image processing to acquire features which can be used for image comparison. Then, optimization algorithms are used to place images into a two-dimensional rectangular similarity grid where similar images are placed closer together. The result of the thesis is an interactive image viewer, designed as a similarity grid. It is implemented in the Python programming language and included in the Orange toolbox as a widget named Image Grid, where it is seamlessly integrated with other widgets in the Image Analytics add-on.

Item Type:

Thesis (EngD thesis)

Keywords:

image similarity grid, interactive image viewer, convolutional neural networks, image embeddings, dimensionality reduction, Orange

Number of Pages:

Language of Content:

Slovenian

Mentor / Comentors:

Name and Surname	ID	Function
prof. dr. Blaž Zupan	106	Mentor

Link to COBISS:

http://www.cobiss.si/scripts/cobiss?command=search&base=51012&select=(ID=1537879235)

Institution:

University of Ljubljana

Department:

Faculty of Computer and Information Science

Item ID:

4154

Date Deposited:

27 Aug 2018 12:32

Last Modified:

06 Sep 2018 10:52

URI:

http://eprints.fri.uni-lj.si/id/eprint/4154

Actions (login required)

View Item