ROC improves liveness, tattoo recognition in facial recognition SDK update

Rank One Computing has enhanced the presentation attack detection and several other algorithms in its facial recognition software development kit.

The new passive, single-frame facial liveness capability delivers a two to five-times reduction in error rate across various spoof mediums and genuine user presentation categories, according to the company announcement. ROC says its biometric PAD accuracy is now up to 99 percent. A further update to ROC’s liveness detection is expected in early-2023.

The accuracy of the SDK’s tattoo recognition algorithm has also been dramatically improved, the company says, by deploying cutting-edge convolutional neural network and machine learning advancements. The improvement is so dramatic, according to the announcement, that the ROC algorithm now delivers accuracy an order of magnitude higher than the leading algorithm in NIST’s 2018 Tatt-e benchmark.

ROC SDK v2.3 also includes overhauled Facial Analytics functions. The company says it can accurately estimate dozens of features in under 50 milliseconds on a single CPU core. The features include mask detection, and compliance checks for ISO/IEC 19794-5 and ICAO 9303 portrait quality. Cropping and background removal features can also be used to make an image ICAO-compliant for facial recognition comparisons.

The SDK’s license plate recognition capabilities have also been upgraded in the new release.

Two ROC algorithms scored notably well in recent NIST testing on facial image quality assessment.

Article Topics

accuracy  |  algorithms  |  biometric identification  |  biometrics  |  facial recognition  |  ISO/IEC 19794-5  |  passive facial liveness  |  presentation attack detection  |  ROC