Advancing the Modalities That Matter Most
ROC SDK 3.14 delivers meaningful improvements across ROC’s core algorithm portfolio for identity, safety, and investigative workflows.
Face Recognition
ROC SDK 3.14 improves accuracy for both Accurate and Very Accurate recognition modes, delivering a 1.22x reduction in errors across our internal datasets. The result is stronger face-matching performance across mission-critical identity workflows.
Age Estimation
#1 U.S. Age Estimation Provider, Ranking #1 in Child Online Safety and Mugshot MAE
Our latest update improves accuracy when face recognition and analytics are used together. By combining top-tier face recognition with #1 NIST-ranked age estimation, ROC delivers more precise, actionable insights for child online safety and related age assurance workflows.
Learn More about the NIST FATE results.
Fingerprint Recognition
ROC SDK 3.14 improves accuracy for controlled fingerprint matching, achieving 2–4% gains in the NIST PFT (1:1) report, along with improved minutiae detection and feature extraction. Our latest IDFP update delivers a 72% reduction in missed identifications at a high-security threshold (1 in 1,000 false negatives).
Latent Fingerprint Recognition
Top Rank-1 Accuracy and Fastest Search Speeds on DoD Dataset
Accuracy is improved for latent fingerprint identification, further advancing ROC’s performance in forensic and investigative environments with industry-leading search speeds that help move cases faster.
Learn More about the NIST ELFT results.
Additional Updates in SDK 3.14
Fingerprint Workflow Improvements
New Merged Template Component APIs give developers more visibility into multi-fingerprint matching, allowing users to deconstruct multi-finger matches into individual component scores and corresponding Template ID pairs. This makes it easier to understand which specific finger drove the match.
Gun Detection
Our latest Gun Detection algorithm strengthens performance and reliability, cutting false rejects by up to 13.2%. This helps ensure critical threats are identified rapidly while maintaining some of the industry’s most stringent false detection controls.
Vehicle Detection and Classification
One model, total visibility. We’ve unified our vehicle architecture to deliver seamless detection across both ground-level and aerial imagery, while expanding classification to include granular subtypes—like sedans and pickups—for more precise forensic filtering.
Fast Object Detection
Our new Fast Detection mode offers a massive throughput boost for person and vehicle tracking, delivering up to a 10x speedup on CPU and 1.2x on CUDA to maximize real-time FPS on resource-constrained hardware and mobile devices.
iOS CoreML Acceleration
Our new CoreML integration brings full hardware acceleration to iOS, delivering lightning-fast face representation and liveness detection, helping improve on-device performance for Apple-based deployments.
Thread Limits Removed
SDK licenses no longer enforce a thread limit. Customers can now take full advantage of available machine resources without artificial constraints.
Future-ready insights.
Straight from the source.
Subscribe for Vision AI insights, product updates, and stories from the front lines of identity and antelligence.