Dataset Information

Face recognition accuracy of forensic examiners, superrecognizers, and face recognition algorithms.

ABSTRACT: Achieving the upper limits of face identification accuracy in forensic applications can minimize errors that have profound social and personal consequences. Although forensic examiners identify faces in these applications, systematic tests of their accuracy are rare. How can we achieve the most accurate face identification: using people and/or machines working alone or in collaboration? In a comprehensive comparison of face identification by humans and computers, we found that forensic facial examiners, facial reviewers, and superrecognizers were more accurate than fingerprint examiners and students on a challenging face identification test. Individual performance on the test varied widely. On the same test, four deep convolutional neural networks (DCNNs), developed between 2015 and 2017, identified faces within the range of human accuracy. Accuracy of the algorithms increased steadily over time, with the most recent DCNN scoring above the median of the forensic facial examiners. Using crowd-sourcing methods, we fused the judgments of multiple forensic facial examiners by averaging their rating-based identity judgments. Accuracy was substantially better for fused judgments than for individuals working alone. Fusion also served to stabilize performance, boosting the scores of lower-performing individuals and decreasing variability. Single forensic facial examiners fused with the best algorithm were more accurate than the combination of two examiners. Therefore, collaboration among humans and between humans and machines offers tangible benefits to face identification accuracy in important applications. These results offer an evidence-based roadmap for achieving the most accurate face identification possible.

SUBMITTER: Phillips PJ

PROVIDER: S-EPMC6004481 | biostudies-literature | 2018 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Face recognition accuracy of forensic examiners, superrecognizers, and face recognition algorithms.

Phillips P Jonathon PJ Yates Amy N AN Hu Ying Y Hahn Carina A CA Noyes Eilidh E Jackson Kelsey K Cavazos Jacqueline G JG Jeckeln Géraldine G Ranjan Rajeev R Sankaranarayanan Swami S Chen Jun-Cheng JC Castillo Carlos D CD Chellappa Rama R White David D O'Toole Alice J AJ

Proceedings of the National Academy of Sciences of the United States of America 20180529 24

Achieving the upper limits of face identification accuracy in forensic applications can minimize errors that have profound social and personal consequences. Although forensic examiners identify faces in these applications, systematic tests of their accuracy are rare. How can we achieve the most accurate face identification: using people and/or machines working alone or in collaboration? In a comprehensive comparison of face identification by humans and computers, we found that forensic facial ex ...[more]

PMID: 29844174

Similar Datasets

Project description:Forensic anthropology and forensic facial image identification are areas with two aspects in common: (i) the use of anthropological knowledge concerning human variation in their analyses and (ii) low numbers of accredited forensic units. While the low numbers are often explained by the uniqueness of human identification cases, given the high level of scrutiny in the forensic sciences, interest in and efforts to achieve accreditation have become increasingly prominent. Therefore, this study aimed to obtain accreditation-relevant information about the working environment in facial image comparison units by surveying facial examiners across Europe. Two surveys were distributed: One was given to participants of a European Commission-funded training course for technical assessors in facial image comparison, and the second, more comprehensive survey, was distributed to members of the European Network of Forensic Science Institutes Digital Imaging Working Group. Thirty-four responses from facial examiners from 16 countries were received. All respondents worked for a governmental organization, nine (26.5%) in accredited units, and 12 (35.3%) had worked as facial examiners for more than 11 years. More than 80% of respondents had an academic background. All examiners from accredited units reported having standard operating procedures, annual Digital Imaging Working Group proficiency testing, and using a standard methodology (compared with 72%, 92%, and 84% from nonaccredited units, respectively). The survey found that working conditions in forensic facial image identification vary among European countries. Some respondents from nonaccredited units reported that their unit had no standard operating procedures, with proficiency tests and intralaboratory validations not performed regularly, and an inconsistently used standard methodology. As these conditions are typically required for successful accreditation, a better understanding of best practice and accreditation requirements in the field is needed. Facilitating interactions between forensic practitioners and quality managers may prove beneficial for future accreditation efforts.Key pointsSurvey of European facial examiners focused on accreditation-relevant topics.Respondents from accredited facial image comparison units differed from those of nonaccredited units in terms of better awareness of standard operating procedures, uptake of intra- and interlaboratory testing, and the consistent use of standard methods and regular method validation.A better understanding of accreditation requirements and best practices recommendations would help to harmonize the practice of the forensic sciences related to human identification.

Dataset Information

Face recognition accuracy of forensic examiners, superrecognizers, and face recognition algorithms.

Publications

Face recognition accuracy of forensic examiners, superrecognizers, and face recognition algorithms.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets