SciTePress - Affiliation Profile

Research.Publish.Connect.

Title:
ISBN:
Year:
Acronym:
Subject:

Advanced Search Proceedings Search

If you're looking for an exact phrase use quotation marks on text fields.

Title:
Author:
Affiliation:
Subject:

Advanced Search Papers Search

If you're looking for an exact phrase use quotation marks on text fields.

Name:
Affiliation:
Country:
Conference:
Subject:

Advanced Search Authors Search

If you're looking for an exact phrase use quotation marks on text fields.

Name:
Country:
Subject:

Advanced Search Affiliations Search

If you're looking for an exact phrase use quotation marks on text fields.

Proceedings

Proceedings Search

Title:
ISBN:
Year:
Acronym:
Subject:

Advanced Search Proceedings Search

If you're looking for an exact phrase use quotation marks on text fields.

Papers

Papers Search

Title:
Author:
Affiliation:
Subject:

Advanced Search Papers Search

If you're looking for an exact phrase use quotation marks on text fields.

Authors

Authors Search

Name:
Affiliation:
Country:
Conference:
Subject:

Advanced Search Authors Search

If you're looking for an exact phrase use quotation marks on text fields.

Advanced Search

Affiliation

Chair of Human-Computer Interaction, Faculty of Mathematics and Computer Science, University of Hagen, Universitätsstraße 47, 58097 Hagen, Germany

Context (please use complete words only)

Field

Sort by

1 Papers

On the Robustness of Convolutional Neural Networks Regarding Transformed Input Images

Topics: Deep Learning; Convolutional Neural Networks

Authors: Frederik Timme , Jochen Kerdels and Gabriele Peters

Affiliation: Chair of Human-Computer Interaction, Faculty of Mathematics and Computer Science, University of Hagen, Universitätsstraße 47, 58097 Hagen, Germany

Digital Object Identifier: 10.5220/0010107203960403

Proceedings of NCTA 2020/IJCCI, ISBN: 978-989-758-475-6

Quick Abstract

Convolutional Neural Networks (CNNs) have become the dominant and arguably most successful approach for the task of image classification since the release of AlexNet in 2012. Despite their excellent performance, CNNs continue to suffer from a still poorly understood lack of robustness when confronted with adversarial attacks or particular forms of handcrafted datasets. Here we investigate how the recognition performance of three widely used CNN architectures (AlexNet, VGG19 and ResNeXt) changes in response to certain input data transformations. 10,000 images from the ILSVRC2012s validation dataset were systematically manipulated by means of common transformations (translation, rotation, color change, background replacement) as well as methods like image collages and jigsaw-like puzzles. Both the effect of single and combined transformations are investigated. Our results show that three of these input image manipulations (rotation, collage, and puzzle) can cause a significant drop in classification accuracy in all evaluated architectures. In general, the more recent VGG19 and ResNeXt displayed a higher robustness than AlexNet in our experiments indicating that some progress has been made to harden the CNN approach against malicious or unforeseen input.