Innovatrics says facial recognition works with masks, adds Trust Score to biometric onboarding toolkit
Innovatrics has added a new Trust Score feature to its Digital Onboarding Toolkit (DOT) to ensure the reliability and trustworthiness of onboarding processes by analyzing the data collected, and also announced that the companies facial biometrics work even with people wearing face masks.
The enhanced data analysis which generates the Trust Score is run by a robustly designed server, Innovatrics says, with AI considering factors including OCR, Two-type face match, age, gender, and liveness, and providing variable thresholds for each to support different use cases, whether fully automated, or assisted onboarding and verification. Depending on whether an onboarding attempt is judged as “Identity Verified,” the client is recommended to “Review Identity,” or the software declares “Identity Not Verified,” fully automated or custom-designed identity verification processes can be applied.
“By adding a trust-based server component, Innovatrics’ digital onboarding platform becomes even more intelligent and secure, capable of comprehensive data analysis to prevent identity theft and fraud,” comments Innovatrics Head of Product Management Daniel Ferak.
DOT provides a platform for simplifying digital onboarding technology deployment, with Innovatrics’ highly-ranked face recognition algorithm and features for auto-capture with visual guidance, liveness detection, facial authentication, and automated OCR for extracting ID data from documents.
The company says its facial recognition recognizes people wearing face masks with only a small drop in accuracy, based on several rounds of testing. Thousands of portraits of faces with and without masks were used to thoroughly test the algorithms capabilities, according to a separate company announcement.
“As soon as masks started becoming commonplace we began collecting testing data to see how face recognition would be affected. What we found out was that even without any changes to the settings, the system is still 95 percent accurate when recognizing a given person in a dataset of hundreds of faces,” explains Metod Rybar, solution manager at Innovatrics.
Innovatrics says its algorithms accuracy is higher than that of human observers in the same conditions, and works with SmartFace and other products without any changes to the way it is implemented. The accuracy can further be improved, however, with a few simple steps. An enrolled picture with a mask, along with one without, with separate thresholds, enables the solution to maintain 99 percent accuracy at 0.01 false non-match rate (FNMR). With a required setting of 0.001, accuracy is still between 90 and 93 percent, the company says. For high-security use cases with requirements of 0.0001 FNMR, accuracy remains above 80 percent.
The company recently announced it had completed an update of Guinea’s electoral register, despite significant challenges, in only six months.