Sight-seeing in the eyes of deep neural networks

None, None; None, None; None, None; None, None; None, None; None, None; None, None

Sight-seeing in the eyes of deep neural networks

Conference Paper (2018)

Author(s)

Seyran Khademi (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Xiangwei Shi (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Tino Mager (TU Delft - Architecture and the Built Environment)

Ronald Siebes (TU Delft - Electrical Engineering, Mathematics and Computer Science, Vrije Universiteit Amsterdam)

Carola Hein (TU Delft - Architecture and the Built Environment)

Victor De Boer (Vrije Universiteit Amsterdam)

Jan Van Gemert (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Research Group

Pattern Recognition and Bioinformatics

Interpretability Classification Visualization Convolutional neural network (CNN) Place recognition

DOI related publication

https://doi.org/10.1109/eScience.2018.00125 Final published version

To reference this document use

https://resolver.tudelft.nl/uuid:c7b3733a-9848-4e55-996f-978b49c98a60

More Info

expand_more

Publication Year

2018

Language

English

Research Group

Pattern Recognition and Bioinformatics

Bibliographical Note

Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Article number

8588744

Pages (from-to)

407-408

Publisher

IEEE

ISBN (electronic)

978-153869156-4

Event

14th IEEE International Conference on eScience, e-Science 2018 (2018-10-29 - 2018-11-01), Amsterdam, Netherlands

Downloads counter

305

Collections

Institutional Repository

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

We address the interpretability of convolutional neural networks (CNNs) for predicting a geo-location from an image. In a pilot experiment we classify images of Pittsburgh vs Tokyo and visualize the learned CNN filters. We found that varying the CNN architecture leads to variating in the visualized filters. This calls for further investigation of the effective parameters on the interpretability of CNNs.

Files

08588744.pdf

(pdf | 0.836 Mb)

- Embargo expired in 01-06-2019

License info not available