Toggle Main Menu Toggle Search

Open Access padlockePrints

The Newcastle University research output collection, currently available on ePrints, will shortly be moving to a new open repository platform, Figshare. To prepare for the data migration we have paused adding new content to ePrints, and will resume once the new repository is launched. During this time you will continue to have access to ePrints (but no new content will appear). We will share updates here when available.

Style augmentation: data augmentation via style randomization

Lookup NU author(s): Phillip Jackson, Dr Amir Atapour AbarghoueiORCiD, Stephen Bonner, Professor Boguslaw ObaraORCiD

Downloads

Licence

This is the authors' accepted manuscript of a conference proceedings (inc. abstract) that has been published in its final definitive form by CVF, 2019.

For re-use rights please refer to the publisher's terms and conditions.

Abstract

We introduce style augmentation, a new form of data augmentation based on random style transfer, for improving the robustness of Convolutional Neural Networks (CNN) over both classification and regression based tasks. During training, style augmentation randomizes texture, contrast and color, while preserving shape and semantic content. This is accomplished by adapting an arbitrary style transfer network to perform style randomization, by sampling target style embeddings from a multivariate normal distribution instead of computing them from a style image. In addition to standard classification experiments, we investigate the effect of style augmentation (and data augmentation generally) on domain transfer tasks. We find that data augmentation significantly improves robustness to domain shift, and can be used as a simple, domain agnostic alternative to domain adaptation. Comparing style augmentation against a mix of seven traditional augmentation techniques, we find that it can be readily combined with them to improve network performance. We validate the efficacy of our technique with domain transfer experiments in classification and monocular depth estimation illustrating superior performance over benchmark tasks.

Publication metadata

Author(s): Jackson P, Atapour-Abarghouei A, Bonner S, Breckon TP, Obara B

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Published

Conference Name: CVPR Workshop 2019

Year of Conference: 2019

Online publication date: 16/06/2019

Acceptance date: 12/11/2019

Date deposited: 06/02/2021

Publisher: CVF

URL: https://openaccess.thecvf.com/content_CVPRW_2019/papers/Deep%20Vision%20Workshop/Jackson_Style_Augmentation_Data_Augmentation_via_Style_Randomization_CVPRW_2019_paper.pdf

Share