Show simple item record

dc.contributor.advisorHuq, Aminul
dc.contributor.advisorReza, Tanzim
dc.contributor.authorAkash, MD Shahadat Hossain
dc.contributor.authorSharife, Shadman Bin
dc.contributor.authorDatta, Bondon
dc.date.accessioned2023-12-07T06:27:15Z
dc.date.available2023-12-07T06:27:15Z
dc.date.copyright2023
dc.date.issued2023-06
dc.identifier.otherID 19101440
dc.identifier.otherID 22241139
dc.identifier.otherID 19101143
dc.identifier.urihttp://hdl.handle.net/10361/21936
dc.descriptionThis thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science, 2023.en_US
dc.descriptionCataloged from PDF version of thesis.
dc.descriptionIncludes bibliographical references (pages 39-40).
dc.description.abstractOne of the most effective biometric tools for securing diverse systems is facial recog- nition technology. Traditional security mechanisms like PINs, passwords, and fin- gerprints have shown to be less effective and dependable than this technology. Sys- tems for surveillance, finance, and security have all made substantial use of facial recognition technology. But facial recognition technology is currently facing a huge challenge from the COVID-19 pandemic. Due to face occlusion brought on by the widespread use of masks, it is now challenging to precisely identify people. This problem has motivated a number of academics to develop facial recognition tech- nology that is more accurate by focusing on hidden facial features. In this paper, we provide a sophisticated method for identifying faces even when there are facial alterations or masks are used. The CASIA, VGG2, LFW Databases are used to help us build our technique, which entails figuring out which facial traits are still discernible even when the face is partially hidden. Convolutional neural networks (CNNs), which are the foundation of our method and are used to extract perti- nent facial information, are based on deep learning techniques. We contrasted our findings with those of other cutting-edge facial recognition systems to assess our strategy. By obtaining more accuracy and speed, our system outperformed other systems. We got 95.85% accuracy in face recognition with the model Inception- ResNetV1 by using ArcFace loss function. Also got 94.28% accuracy on the same model by using Triplet loss function. We worked on the never before worked model for face recognition which is SE-ResNeXt-101. We also got 93.41% accuracy on that model with ArcFace loss function. That indicates that our research underlines the significance of creating environmental change-resistant facial recognitionen_US
dc.description.statementofresponsibilityMD Shahadat Hossain Akash
dc.description.statementofresponsibilityShadman Bin Sharife
dc.description.statementofresponsibilityBondon Datta
dc.format.extent40 pages
dc.language.isoenen_US
dc.publisherBrac Universityen_US
dc.rightsBrac University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.
dc.subjectFacial recognitionen_US
dc.subjectFace detectionen_US
dc.subjectConvolution Neural Networksen_US
dc.subjectDeep Neural Networken_US
dc.subjectInceptionResNetV1en_US
dc.subjectArcFace loss functionen_US
dc.subjectSE-ResNeXt-101en_US
dc.subjectTriplet loss functionen_US
dc.subject.lcshFacial expression.
dc.subject.lcshArtificial intelligence
dc.subject.lcshNeural networks (Computer science)
dc.titleObscure face recognition using deep neural networksen_US
dc.typeThesisen_US
dc.contributor.departmentDepartment of Computer Science and Engineering, Brac University
dc.description.degreeB.Sc. in Computer Science and Engineering


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record