dc.contributor.author | Sáez-Trigueros, Daniel | |
dc.contributor.author | Meng, Li | |
dc.contributor.author | Hartnett, Margaret | |
dc.date.accessioned | 2020-12-18T00:11:30Z | |
dc.date.available | 2020-12-18T00:11:30Z | |
dc.date.issued | 2021-02-01 | |
dc.identifier.citation | Sáez-Trigueros , D , Meng , L & Hartnett , M 2021 , ' Generating photo-realistic training data to improve face recognition accuracy ' , Neural Networks , vol. 134 , pp. 86-94 . https://doi.org/10.1016/j.neunet.2020.11.008 | |
dc.identifier.issn | 0893-6080 | |
dc.identifier.uri | http://hdl.handle.net/2299/23612 | |
dc.description | ©2020 Elsevier Ltd. All rights reserved. This is the accepted manuscript version of an article which has been published in final form at https://doi.org/10.1016/j.neunet.2020.11.008 | |
dc.description.abstract | Face recognition has become a widely adopted biometric in forensics, security and law enforcement thanks to the high accuracy achieved by systems based on convolutional neural networks (CNNs). However, to achieve good performance, CNNs need to be trained with very large datasets which are not always available. In this paper we investigate the feasibility of using synthetic data to augment face datasets. In particular, we propose a novel generative adversarial network (GAN) that can disentangle identity-related attributes from non-identity-related attributes. This is done by training an embedding network that maps discrete identity labels to an identity latent space that follows a simple prior distribution, and training a GAN conditioned on samples from that distribution. A main novelty of our approach is the ability to generate both synthetic images of subjects in the training set and synthetic images of new subjects not in the training set, both of which we use to augment face datasets. By using recent advances in GAN training, we show that the synthetic images generated by our model are photo-realistic, and that training with datasets augmented with those images can lead to increased recognition accuracy. Experimental results show that our method is more effective when augmenting small datasets. In particular, an absolute accuracy improvement of 8.42% was achieved when augmenting a dataset of less than 60k facial images. | en |
dc.format.extent | 9 | |
dc.format.extent | 6750526 | |
dc.language.iso | eng | |
dc.relation.ispartof | Neural Networks | |
dc.subject | Image generation | |
dc.subject | Generative adversarial learning | |
dc.subject | Face and gesture recognition | |
dc.subject | Machine Learning | |
dc.subject | Machine learning | |
dc.subject | Artificial Intelligence | |
dc.subject | Computer Vision and Pattern Recognition | |
dc.subject | Cognitive Neuroscience | |
dc.title | Generating photo-realistic training data to improve face recognition accuracy | en |
dc.contributor.institution | Department of Engineering and Technology | |
dc.contributor.institution | School of Physics, Engineering & Computer Science | |
dc.contributor.institution | Communications and Intelligent Systems | |
dc.contributor.institution | Centre for Engineering Research | |
dc.description.status | Peer reviewed | |
dc.date.embargoedUntil | 2021-11-27 | |
dc.identifier.url | http://www.scopus.com/inward/record.url?scp=85097342288&partnerID=8YFLogxK | |
rioxxterms.versionofrecord | 10.1016/j.neunet.2020.11.008 | |
rioxxterms.type | Journal Article/Review | |
herts.preservation.rarelyaccessed | true | |