Publicação
Deep Learning-based Facial Detection and Recognition in Still Images for Digital Forensics
| dc.contributor.author | Domingues, Patrício | |
| dc.contributor.author | Rosário, Alexandre Frazão | |
| dc.date.accessioned | 2026-03-06T12:42:39Z | |
| dc.date.available | 2026-03-06T12:42:39Z | |
| dc.date.issued | 2019-08-26 | |
| dc.description.abstract | Smartphones and cheap storage have contributed to a deluge of digital photos. Digital forensic analysis often include the need to process large volumes of digital photos found on devices. Sometimes, this is done either to detect or confirm the ownership of the device or to determine whether the owner of the device has some acquaintance of interest in the case. In this paper, we present the Face Detection and Recognition in Images (FDRI) open source software, and its integration as a module for the digital forensic software Autopsy. FDRI aims to semi-automate the detection of faces in digital photos, flagging photos where at least one face is detected. FDRI software also performs face recognition, searching for the existence of given individual(s) in still photos of the forensically examined devices. For both the detection and recognition of faces, FDRI resorts to deep learning-based algorithms available within the dlib machine learning toolkit. In experimental assessments, FDRI yielded an average precision of 99.46% face detection and 98.10% for face recognition, when dealing with the restrained LFW dataset. For unrestrained real world photos, FDRI achieved a precision of 97.67% for face detection and 81.82% for face recognition. Performance-wise, this study confirms the importance of a fast GPU for fast face detection and recognition, with an NVidia GTX 1070 being roughly three times faster than a GTX 750 Ti, and in certain cases, up to 35× faster than the CPU version. | eng |
| dc.description.sponsorship | This work was partially supported by FCT and Instituto de Telecomunicações under project UID-EEA-50008-2013 and by DEI-ESTG Polytechnic Institute of Leiri | |
| dc.identifier.citation | Domingues, Patrício & Frazão, Alexandre. (2019). Deep Learning-based Facial Detection and Recognition in Still Images for Digital Forensics. 1-10. 10.1145/3339252.3340107. | |
| dc.identifier.doi | 10.1145/3339252.3340107 | |
| dc.identifier.isbn | 978-1-4503-7164-3 | |
| dc.identifier.uri | http://hdl.handle.net/10400.8/15798 | |
| dc.language.iso | eng | |
| dc.peerreviewed | yes | |
| dc.publisher | ACM | |
| dc.relation.hasversion | https://dl.acm.org/doi/10.1145/3339252.3340107 | |
| dc.relation.ispartof | Proceedings of the 14th International Conference on Availability, Reliability and Security | |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | |
| dc.subject | face detection | |
| dc.subject | face recognition | |
| dc.subject | digital forensics | |
| dc.subject | deep learning | |
| dc.title | Deep Learning-based Facial Detection and Recognition in Still Images for Digital Forensics | eng |
| dc.type | conference paper | |
| dspace.entity.type | Publication | |
| oaire.citation.conferenceDate | 2019-08 | |
| oaire.citation.conferencePlace | Canterbury | |
| oaire.citation.title | ARES '19: Proceedings of the 14th International Conference on Availability, Reliability and Security | |
| oaire.version | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |
| person.familyName | Domingues | |
| person.givenName | Patrício | |
| person.identifier.ciencia-id | AA15-6185-C477 | |
| person.identifier.orcid | 0000-0002-6207-6292 | |
| person.identifier.scopus-author-id | 13411315400 | |
| relation.isAuthorOfPublication | b88ada5f-0d8b-4e55-ab0a-62aa82ea1388 | |
| relation.isAuthorOfPublication.latestForDiscovery | b88ada5f-0d8b-4e55-ab0a-62aa82ea1388 |