The SAFR Platform from RealNetworks: An AI-Powered Solution for Secure Facial Recognition Enabling New Levels of Visibility

The Safr Platform From Realnetworks

TL; DR: The SAFR platform from RealNetworks is a premier facial recognition solution for live video. Powered by AI, the solution is optimized for accuracy and performance through a marathon-like approach to product development centered on continuous improvement. Designed to accommodate a wide range of environments, from enterprises to K-12 schools, SAFR is providing new safety opportunities to customers worldwide.

If you have listened to music or watched a video online in the last decade, chances are you’ve used RealPlayer, one of the first applications for streaming media over the internet.

The software, formerly known as RealAudio Player, was released in 1995 by RealNetworks. Today, as a multimillion-dollar company publicly traded on NASDAQ, RealNetworks remains at the forefront of innovation on the internet.

“Our 25-year legacy as a pioneer in the video space has led us to create the world’s foremost platform for facial recognition in live video,” said Dan Grimm, Vice President and General Manager of Computer Vision at RealNetworks. “Secure, Accurate Facial Recognition (SAFR) is an AI-powered product that provides next-generation visibility and situational awareness for security professionals.”

SAFR logo

SAFR is an AI-powered facial recognition solution designed to strengthen security.

Unlike other facial recognition platforms, SAFR was designed to accurately identify faces that are in motion, under poor lighting conditions, partially obscured, or misaligned — real-life characteristics that cameras might capture in a natural setting.

SAFR also employs an integrated set of technologies to combine high performance with a flexible architecture. This approach allows the forward-thinking company to support a variety of practical use cases at enterprises and K-12 schools alike, from secure access and gateway safety to VIP loyalty and venue monitoring.

Ultimately, RealNetworks is committed to building the world’s most trusted facial recognition platform through SAFR, leveraging ongoing improvements in both artificial intelligence and machine learning to continually raise the bar in terms of accuracy and performance.

Using AI to Make Sense of the World Around Us

As much as we fear a technological takeover of sci-fi movie proportions, we have to face the facts: Cameras have outnumbered humans on this earth. The current world population is estimated at more than 7.7 billion, and there are approximately 14 billion cameras in the world today. By 2022, that number is predicted to balloon to 45 billion. With nearly six cameras for every person on earth, many humans will never view the majority of the visuals captured by these devices — there is simply too much data for the human brain to process.

“Instead, artificial intelligence will come to the edge and make sense of all of that visual information,” Dan said. “With SAFR, we will enable customers in a variety of verticals in the security space — home, residential, enterprise, airports, hospitals, stadiums — to leverage our platform to understand the world around them.”

SAFR facial recognition examples

SAFR helps maintain persistent awareness to keep people safe.

RealNetworks offers a software development kit (SDK) for SAFR that will enable manufacturers of IoT devices in the edge computing space to leverage its world-class facial recognition technology through live video. This will enable, for example, the head of security at a hospital or stadium to derive new capabilities and insight from hundreds of live video feeds in a way that was humanly impossible in the past.

“Device manufacturers can license our technology to bring truly world-class computer vision and facial recognition to the edge,” Dan said. “We have SDKs that will run on devices such as video conferencing cameras, home security cameras, mobile devices, refrigerators, drones, cars, etc.”

A High Level of Accuracy and Low Propensity for Bias

Dan told us that RealNetworks has observed dramatic advances in deep learning with convolutional neural networks (a type of artificial neural network used to analyze images) over the last few years, and has worked to stay one step ahead in the quickly evolving space.

When assessed against the benchmark of Labeled Faces in the Wild (LFW), a database of photographs developed by the University of Massachusetts to study unconstrained facial recognition under real-world conditions, SAFR demonstrated 99.87% accuracy. In addition, RealNetworks continually submits its algorithms to the National Institute of Standards and Technology (NIST) for performance evaluations.

NIST’s test results from April 2019 ranked the SAFR algorithm as both the fastest and most compact among algorithms for wild images, with a false non-match rate (the rate at which an algorithm mistakenly identifies two images of the same individual as different) of less than 0.025. In terms of template extraction — the process of creating a facial signature — NIST ranked SAFR’s algorithm among the top seven worldwide.

“We are committed to creating a computer vision platform that is truly excellent, exhibiting high levels of accuracy and low levels of bias in terms of variances in skin tone and gender — these are things that we are continuously improving upon,” Dan said. “We believe this technology offers tremendous benefits but has to meet a certain bar of excellence for us to retain customer trust.”

Increase School Safety for Free in K-12 Schools

RealNetworks introduced SAFR to the market in 2018 as a free download for K-12 schools in the United States and Canada looking to improve security and convenience for trusted members of the community. After guest complete a registration process on a tablet with staff approval, the technology will match faces in real time to streamline the entry and check-in process.

“As a parent or a staff member, you can register yourself in the lobby of the school,” Dan said. “The next time you come to the front gate of the campus, instead of standing there and hitting a buzzer that someone may or may not answer, with or without knowledge of who is there, a camera will recognize you and automatically unlock the door.”

SAFR at a school

The AI-powered solution identifies faces in real-world conditions.

In the same way that the technology makes security-conscious campuses accessible to trusted members of the community, it may also be configured to recognize persons of concern — such as expelled students or former employees — from within and outside of the school. In this way, the platform adds real value for North American schools, which have been plagued with deadly violence over the past few decades.

“We help our customers, particularly in the security space, see what matters in the real world,” Dan said. “We have begun with facial recognition, but we’ve also recently launched with person detection and will continue to expand our vision offering to address the needs of our customers for visual intelligence in the spaces they care about.”

A Marathon Approach to Improving Accuracy

Dan said building highly accurate facial recognition software is a marathon, not a sprint. In addition to listening to feedback from the education and enterprise security communities, RealNetworks strives to continuously improve the accuracy and speed of its algorithms to remain at the forefront of the industry.

“The reality is that there are many facial recognition products in the market, and that number grows every quarter, but they vary dramatically in terms of performance,” he said.

Ultimately, the mission at SAFR and RealNetworks is to build the world’s most trusted visual recognition platform. “In order to satisfy that responsibility, we believe we have to design and develop a truly excellent product,” Dan said.

Advertiser Disclosure is a free online resource that offers valuable content and comparison services to users. To keep this resource 100% free, we receive compensation from many of the offers listed on the site. Along with key review factors, this compensation may impact how and where products appear across the site (including, for example, the order in which they appear). does not include the entire universe of available offers. Editorial opinions expressed on the site are strictly our own and are not provided, endorsed, or approved by advertisers.

Our Editorial Review Policy

Our site is committed to publishing independent, accurate content guided by strict editorial guidelines. Before articles and reviews are published on our site, they undergo a thorough review process performed by a team of independent editors and subject-matter experts to ensure the content’s accuracy, timeliness, and impartiality. Our editorial team is separate and independent of our site’s advertisers, and the opinions they express on our site are their own. To read more about our team members and their editorial backgrounds, please visit our site’s About page.