RealNetworks
RNWK)

On the fifth of September 1995, a baseball fan could—for the first time—listen to the live broadcast of a baseball game between the New York Yankees and Seattle Mariners through a computer. This was only possible due to the efforts of a company named RealNetworks and its RealAudio software that, at the time, emerged as a pioneer to ignite the streaming revolution. Quickly becoming a mogul in the media streaming industry in the 90s, RealNetworks strode forward to create the massively popular media player, RealPlayer, and a video/image sharing and editing app called RealTimes. RealNetworks was consequently empowered to leverage this long video/image technology experience and forge the Machine learning based AI facial recognition software: SAFR

The Next Generation of Surveillance.

SAFR (Secure Accurate Facial Recognition) is a visual intelligence platform powered by machine learning (ML), intuitively designed to revolutionize biometrics, combat fraud, and bolster security. With this incredibly sophisticated piece of innovation, RealNetworks is redefining facial recognition technology in today’s hyper-connected, globalized marketplace. The solution shows much promise in the areas of threat detection and behavioral observations as well, offering several applications for both commercial and government entities to ensure peoples’ safety. “According to a recent survey by the National Institute of Standards and Technology (NIST), SAFR is one of the fastest algorithms in terms of detection speed and also requires minimal storage space. It is one the world’s best facial recognition solutions,” says Noriaki Takamura, APAC Vice president of RealNetworks. NIST’s test for the efficacy of facial recognition software is considered the benchmark worldwide; it stringently checks for accuracy, performance, and bias thrice a year. And, in 2019, over the course of three months, SAFR only got better, displaying a thirty percent increase in speed. The solution’s ability to carry out five hundred recognitions per second and a load balance of this massive size over multiple streams easily overtakes the current standard achieved in the industry.



Gleaning data from one of the largest image databases in the world, SAFR has an extremely high detection accuracy of 99.87 percent and can even recognize individuals using face masks, with a staggering 99.5 percent accuracy. Moreover, with data from users across regions such as Africa and Southeast Asia, the software is optimized to detect various skin tones and complexions. “A defining edge that SAFR has over competitors is the solution’s ability to precisely detect skin tone and user liveness,” adds Takamura.

The latest version of SAFR has further been engineered to passively detect liveness for biometric authentication solutions.

A rise in cases of video conference hacking, spoofing attempts, and other similar cases of fraud have prompted companies to seek out solutions that can passively verify the authenticity of the user. Provided with a still image or a screengrab from a video, most facial recognition software fails to distinguish between a real face and a portrait. The SAFR algorithm completely circumvents this shortcoming by utilizing a unique, multi-faceted approach to analyzing texture and context to successfully detect static images and fabricated video clips. Unlike similar solutions that rely on projected light, infrared or specialized stereoscopic/depth-sensing cameras that are expensive, SAFR can be used to analyze video streams from internet protocol (IP) cameras embedded in access control terminals. To that end, RealNetworks recently entered a partnership with Axis Communications, a Swedish video surveillance and camera manufacturer, and has been able to implement SAFR Indie into two camera models—the P3255-LVE and the Q1615 MK III. This integration saves the customer from investing in expensive edge computing hardware, thereby considerably reducing the total cost of ownership.

An Incredibly Intelligent Video Surveillance Solution

SAFR is designed to reign supreme in the realms of access control, identity identification, and automated surveillance. And to elaborate on the same, Takamura recalls an interaction with the Kolkata police where SAFR’s facial recognition prowess proved to be highly reliable in arresting suspected felons off the city’s streets. The police force partnered with RealNetworks to embed SAFR into the massive network of several IP cameras spread throughout the metropolis and proceeded to feed SAFR’s algorithm with criminals’ data stored on the cloud. During a large book fair in 2019, the cameras notified authorities of individuals who matched the descriptions provided on various watchlists, which then led to the arrests of four criminals.

In another instance, RealNetworks was implemented by a large Japanese company that ran a thermal power plant. Owing to the fact that such facilities span significant distances to cover on foot, shuttle services were provided to employees and visitors at the entrance. Normally carrying close to thirty people, the shuttle required that every passenger be registered at the reception for security purposes. This process, however, was lengthy and time-consuming. After SAFR was integrated into the 5G mobile phone for security guard and IT infrastructure, security guards gets in to shuttle to check authorized and unauthorized passengers by the 5G phone in few min.

In populated and demographically diverse regions like India, SAFR’s facial recognition software serves as a much more effective and cost-efficient alternative to prevalent identity identification solutions such as fingerprint and iris verification. Moreover, the pandemic has spurred organizations to implement non-contact identification software. To further improve the capabilities of SAFR, SAFR has done integration with key video management systems (VMS) such as Milestone, Genetec, Geutebruck, Digifort, Avigilon and server solutions providers, manufacturers (Dell, HPE, Supermicro, NVIDIA, etc.), and system integration partners. SAFR recognition speed of <100ms is one of the world’s fastest algorithms in NIST and its new version 3.6 consists of liveness detection as well. RealNetworks recently signed a reseller agreement with NTT Docomo, Japan’s largest mobile carrier, to offer SAFR’s innovative AI-based facial recognition software. SAFR is expected to provide new options for Docomo’s customers by offering reliable and high-speed AI-facial recognition capabilities in combination with its existing product portfolio. “We plan to develop a secure ecosystem that not only helps clients in their surveillance and authentication endeavors but also stays in line with modern data protection regulations,” concludes Takamura.