SAFR for Security

A new level of visibility and situational awareness with exceptionally accurate, AI-powered facial recognition.

“Facial recognition has long been promised as a boon to security professionals, but too often it has fallen short on performance. SAFR for Security is powerful because it works — it’s fast, accurate, and reliable in a surprising number of adverse conditions. We’re delighted to bring this solution to our clients to improve security and situational awareness.”

Dae-Jun Lee, Manager, SK Telecom

SAFR for Security is available as a standalone solution or integrated with a VMS

SAFR for VMS features

Alarms & Notifications

Security teams can customize real-time alerts and be instantly notified when persons of interest enter or leave a monitored area. No matter the use case – multi-factor authentication, building system integration, unobtrusive monitoring, threat detection or more – notifications can be further customized to initiate powerful security responses.

Automatic Bookmarks

Teams can create automatic bookmarks for a variety of conditional scenarios. Bookmarks and searchable metadata enable more efficient investigative and forensic work with recorded video.

“The SAFR for Security AI facial recognition platform provides an excellent complement to our AI video analytic security solutions. The integration is easy, and we’re excited by the opportunities to deliver the combined real-time video analytics plus SAFR solutions to our customers in Korea and overseas.”

Young Hoon Bae, CEO, IVS Inc.

Fast & Accurate

In the world of security, reaction time is everything. But delivering rapid facial recognition results without accuracy would return unacceptable errors, like granting access to a threat or denying access to a VIP. SAFR’s algorithm has been designed for a superior balance of speed and accuracy.

SAFR’s 99.87 percent accuracy is balanced by industry-leading performance that delivers results 3-5 times as fast as competing facial recognition algorithms. In the July 2019 National Institute of Standards and Technology (NIST) test results, the SAFR algorithm tested as both the fastest and most compact amongst algorithms for wild images with less than 0.0335 False Non-Match Rate (FNMR). Compared to the other high-accuracy algorithms, SAFR is 2x faster than average, and 35 percent smaller than the second-smallest algorithm. SAFR was the fastest algorithm tested in the April 2019 NIST results and its speed improved by 30% in the 2019 July results. This rapid performance allows SAFR to sample video image frames considerably faster than competing algorithms. Increased sampling increases accuracy. The superior accuracy achieved by SAFR is also a result of a greater diversity in the data set and the use of only real faces (vs. simulated faces) to train the algorithm. The SAFR database is optimized for rapid matching against millions of records in near real time, making identification extraordinarily fast. With edge intelligence, SAFR detects a face moving through live video, and then recognizes it, in under 100 milliseconds.

SAFR also has one of the lowest rates of bias with respect to skin tone and gender, according to NIST tests of more than 100 algorithms. SAFR performs consistently across a range of skin tones due to its massive and highly diverse global training set.

Secure

SAFR is built for privacy. All facial signatures and facial images in transit or at rest use AES-256 encryption, and no data is passed over the internet if run on premises. Privacy is thoughtfully addressed from a full suite of administrative tools that include easy opt-in/opt-out capabilities to Implementation Best Practices.

Distributed & Flexible

SAFR’s distributed architecture allows for efficient bandwidth consumption, reducing the roundtrip latency, even over 3G networks. The low-bandwidth requirements, edge processing, efficient algorithm, off-the-shelf hardware, and ability to leverage inexpensive GPUs deliver significant cost savings.

The highly adaptable SAFR platform is ideally suited for a range of real-world security use cases where reliable, exceptionally accurate industrial-grade facial recognition is required. Fully configurable alerts enable custom actions based on recognition events, or data exporting for insights and reports. SAFR enables new applications for security, convenience, and live analytics.

Deployment Neutral

Unlike other solutions requiring full video uploads to the cloud – incurring significant bandwidth costs and expensive overhead – the SAFR platform can be deployed on a single computer to monitor a handful of IP cameras, or scaled to thousands of cameras to meet the challenges security professionals face every day. The distributed architecture enables optimized performance, hosted on premises or in the cloud.

Actionable

The SAFR platform makes it easy to assign people and roles, view security cameras through the SAFR interface, and configure powerful custom actions and alarms based on recognition events: unlocking a door, turning on lights, sounding an alarm, denial of entry, logging demographic information, generating reports, notifying police, deploying security guards, initiating a building lockdown – any security response is possible. SAFR also provides actionable data for live analytics of traffic volumes, demographic composition, dwell times, and data exports.