Comparative performance of face biometric verification system with multi selfie devices

This paper addresses the research question: "Can face images from selfie devices be used instead of images from DSLR cameras in facial biometric verification systems?" To investigate this, we propose a method for comparing the performance of a facial biometric verification system using multiple selfie devices. Experiments were conducted on facial images captured from three different devices: a DSLR camera, an iPhone, and an Android mobile phone, with a dataset of 1,000 front-view individuals. The Neurotechnology Face Verification system, ranked the NIST Face Recognition Technology Evaluation, was used for verification. The images from these devices were processed for 1:1 verification, considering both genuine and impostor cases. Performance metrics, including FNMR, FMR, FRR, FAR, and DET, were analyzed to evaluate system accuracy. Overall, the performance values for images from all three devices were nearly identical, suggesting that images from a DSLR, iPhone, and Android mobile phone can be interchangeably used for facial biometric verification systems.