Rank One updates SDK for enhanced facial recognition accuracy and speed

Rank One Computing has announced an update to its ROC face recognition SDK which it says substantially improves accuracy, with error rates as much as 40 percent lower for the new version 1.20, according to the company.

The embedded algorithm designed for resource-constrained applications such as those on mobile devices or using high-throughput video and included in the ROC SDK features the same accuracy, but 10 percent faster performance, Rank One says.

The upgrade also includes an overhaul of the SDK’s face detection algorithm that boosts both accuracy and speed, the “Ethnicity” classifier has become a “Geographic Origin” classifier, and faces in profile (Yaw pose angle greater than 60 degrees) can now be manually enrolled. A new classifier was added to detect facial drawings, and new licensing options have been made available. New API wrappers for Python 3 and Lua are included, and the optional CUDA GPU enrollment pipeline has been made more efficient.

Rank One also says its v1.20 algorithm, referenced as “rankone-008,” was one of the only submissions to score better than the median in all measured categories in the latest NIST Face Recognition Vendor Test (FRVT) Ongoing report.

A previous version of the ROC SDK also scored well in DHS’ Biometric Technology Rally earlier this year.

Article Topics

accuracy  |  biometrics  |  face detection  |  facial recognition  |  Rank One Computing  |  SDK