Rank One is pleased to announce the release of version 1.16 of the ROC SDK, which includes another round of significant feature enhancements to the world’s most versatile face recognition solution.
Several accuracy improvements have been made to the SDK in this release. First, version 1.16 provides across-the-board precision improvements for all capture scenarios (constrained cooperative, unconstrained cooperative, and unconstrained uncooperative). These accuracy improvements are on top of the major improvements released in version 1.15. Second, face recognition accuracy is now more equally balanced across different racial and gender cohorts. Finally, major enhancements to the face detection and recognition framework now allow users to perform recognition across all facial pose angles, including full-pose (-90° to 90° yaw) face detection and profile to frontal face matching. Adding “full-pose” face recognition capabilities is the first step in a series of scheduled enhancements that will enable our users to achieve successful facial matches despite extreme facial pose angles and low quality capture environments.
Perhaps the most significant improvements in this release relate to algorithmic efficiency. Previous versions of the SDK were already proven to be the world’s most efficient face recognition solution in terms of comparison speed and enrollment speed, per testing from the National Institute of Standards and Technology (NIST). In version 1.16, template size has been reduced by 60%, down to 133 bytes. The ROC SDK’s small template size, unparalleled comparison and enrollment speeds and robust accuracy are the result of the Rank One team’s relentless research, and in turn yield significant hardware cost savings for customers and enable on-edge applications that would otherwise be impossible.
Rank One is also excited to release major enhancements to its passive liveness/anti-spoof algorithm. Rank One’s patent-pending technique requires no special hardware and is effective at deterring printed photo and digital screen replay attacks using a single image capture, meaning powerful liveness/anti-spoofing can be performed passively, without requiring multiple images, multiple poses, user response or specialized hardware. This liveness solution performs well on both fixed-focus and auto-focus cameras, thereby covering both front-facing and rear-facing cameras on mobile devices and addressing both selfie and access control use cases.
As always, there are several enhancements to the usability of the SDK. Ease-of-integration is one of the key intangible reasons the size of Rank One’s customer base continues to accelerate. Rank One is committed to continued excellence on the ease-of-integration front, whether it is through our stable well-thought-out API, support for multiple programming languages, command line applications, or responsive customer support provided by the same engineers who develop the technology.