Please use this identifier to cite or link to this item: http://drsr.daiict.ac.in//handle/123456789/1016
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorBhilare, Shruti
dc.contributor.advisorMitra, Suman K.
dc.contributor.authorSherasia, Kehkasha
dc.date.accessioned2022-05-06T17:53:05Z
dc.date.available2023-02-24T17:53:05Z
dc.date.issued2021
dc.identifier.citationSherasia, Kehkasha (2021). Presentation Attack Detection in Face Recognition System. Dhirubhai Ambani Institute of Information and Communication Technology. xi, 43 p. (Acc.No: T00951)
dc.identifier.urihttp://drsr.daiict.ac.in//handle/123456789/1016
dc.description.abstractOne of the quickest, most precise and easily available biometric recognition system is Face recognition. These systems have wide range of uses like phone verification, payment method security checks, border control and surveillance. These systems however, are subject to a variety of spoof attacks which are also known as presentation attacks. Hence it is necessary that efficient face anti-spoofing methods are developed. Here in our research work, we use CNN (convolutional neural network) model along with class activation maps for detection of spoof attacks. This approach helps us extract local information and CNN aids in the development of a robust model. We have performed experiments on challenging benchmark dataset OULU-NPU. We use Class Activation Map for the classification task. We achieved an accuracy of 95% using the proposed approach.
dc.subjectPresentation Attack
dc.subjectFace Anti-Spoofing
dc.subjectClass Activation Map
dc.subjectMultichannel CNN
dc.classification.ddc006.42 SHE
dc.titlePresentation Attack Detection in Face Recognition System
dc.typeDissertation
dc.degreeM. Tech
dc.student.id201911024
dc.accession.numberT00951
Appears in Collections:M Tech Dissertations

Files in This Item:
File Description SizeFormat 
201911024_Image_Spoofing_thesis - Shruti Bhilare.pdf
  Restricted Access
4.25 MBAdobe PDFView/Open Request a copy


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.