ROC SDK version 1.17 is now available. With this new release users will receive more accuracy enhancements, faster processing speeds, and usability improvements.
The biggest improvement with version 1.17 is another order-of-magnitude reduction in false positive error rates when operating on unconstrained, or “Wild”, imagery. These significant upgrades in accuracy will be particularly noticeable to integrators that process imagery with variations in facial pose and environmental illumination.
In addition to large accuracy improvements, processing speeds for v1.17 are roughly 10% faster than v1.16’s industry leading speed and efficiency.
The final major improvement delivered in v1.17 is a significant extension to the video processing API, which was enabled by the Rank One team porting a large amount of the source code from our Explore GUI system into the SDK and making these features accessible through a well thought-out API. These enhanced features will reduce the overhead required to develop tracking and monitoring applications using the ROC SDK.
When asked about the achievements delivered in ROC SDK v1.17, Chief Engineer Scott Klum said, “Rank One’s core algorithms have always been on the cutting edge of accuracy and efficiency. With version 1.17, we had a number of breakthroughs that simplified algorithm architecture while, in keeping with Occam’s razor, providing significant accuracy gains. These improvements give our users continued access to a technology so efficient that it can run on an embedded device, with accuracy comparable to (and in most cases superior to) competitors’ technology that must be run on a server.”
Rank One is not letting up on algorithm R&D anytime soon. Users who have become accustomed to major quarterly enhancements to the SDK can expect another big boost in accuracy that is already underway and scheduled for release in mid-December.