IIIT Hyderabad Publications |
|||||||||
|
Universal Material Translator: Towards Spoof Fingerprint GeneralizationAuthors: Sri Aurobindo Munagala,Additya Popli,Tarang Chugh,Anoop Namboodiri,Anil K Jain Conference: The 12th IAPR International Conference on Biometrics (ICB 2019 2019) Location Crete, Greece Date: 2019-07-04 Report no: IIIT/TR/2019/83 AbstractSpoof detectors are classifiers that are trained to dis-tinguish spoof fingerprints from bonafide ones. However,state of the art spoof detectors do not generalize well onunseen spoof materials. This study proposes a style transferbased augmentation wrapper that can be used on any ex-isting spoof detector and can dynamically improve the ro-bustness of the spoof detection system on spoof materialsfor which we have very low data. Our method is an ap-proach for synthesizing new spoof images from a few spoofexamples that transfers the style or material properties ofthe spoof examples to the content of bonafide fingerprintsto generate a larger number of examples to train the clas-sifier on. We demonstrate the effectiveness of our approachon materials in the publicly available LivDet 2015 datasetand show that the proposed approach leads to robustness tofingerprint spoofs of the target material. Full paper: pdf Centre for Visual Information Technology |
||||||||
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved. |