The NIST has found that facial recognition algorithms have improved their ability to detect identities even when the individual is wearing a mask. Since the beginning of the pandemic, numerous studies have been conducted to test the effect of masked faces on algorithms, as facial recognition software pre-pandemic was unable to identify masked individuals. According to the NIST, some error rates decreased as much as by a factor of 10 between pre and post COVID algorithms.
NIST stated that in the best cases, software algorithms are making errors only 2.4-5% of the time on masked faces, which is comparable to the performance of the technology in 2017 with clear, full-face photos. The study accumulated 65 newly submitted algorithms that were tested, offering results for 152 total algorithms to present a full picture of the technology’s capabilities.
Read More: Face recognition software making progress at recognizing masked faces
