Face masks have become known to stop the spread of Covid-19. But it can also make it harder for facial-recognition software for identifying you. This is an important finding of a new report released by NIST, which talks about the accuracy of the facial-recognition algorithms submitted by agencies and researchers to labs, a branch of the U.S Department of Commerce.
What NIST says?
Under the circumstances, the National Institute of Standards and Technology states that the failure rate for the best facial recognition systems is about 0.3%, although the study has found disparities across gender, race, and age.
They tested the face-recognition software by drawing digital masks in the faces of cross-border photographs and then compared those images to another database of people exposed to visas and other immigration benefits. The organization says that it scanned 6.3 million images of about 1 million people who are using 89 algorithms provided by technology companies and academic lab providers.
Findings
- Masked images frequently disable algorithms to process any face. Facial recognition algorithms usually work by measuring facial features – their size and distance from each other. As per the report, with masks, the algorithm cannot find out enough about any facial features to make effective comparisons in the first place.
- The study looked at three levels of nasal coverage: low, medium, and high – specifying that degradation occurs with greater nasal coverage.
- Although false negatives have increased, and false positives have decreased steadily. A moderate reduction in the false-negative rate shows that inclusion with masks does not weaken the aspect of protection.
Automatic Facial Recognition Is More Accurate Than the Current One
- The accuracy of the measurement of visual human passport inception is notoriously low. It is within the range of 80% or less which is determined by a person.
- According to NIST, Algorithm performance for high performers across the board is 20 times better than skilled professional testers.
The research from the NIST has shown that more facial features are covered, the harder it is to identify a person under the mask.
Want to publish your own articles on DistilINFO Publications?
Send us an email, we will get in touch with you.