TL; DR: The SAFR platform from RealNetworks may be a premier facial recognition choice for live video. Electric by AI, the solution is optimized meant for accuracy and performance from a marathon-like approach to product development devoted to continuous improvement. Designed to accommodate several environments, from enterprises that will K-12 schools, SAFR is providing new safety opportunities that will customers worldwide.
For people with listened to music or simply watched a video online in the last decade, chances are you’ve put to use RealPlayer, one of the earliest applications for streaming media on the internet.
The software, formerly also known as RealAudio Player, was revealed in 1995 by RealNetworks. At this time, as a multimillion-dollar provider publicly traded on NASDAQ, RealNetworks remains at any forefront of innovation via the internet.
“Our 25-year legacy as the pioneer in the video tutorial space has led us to develop the world’s foremost base for facial recognition during live video, ” says Dan Grimm, Vice Director and General Manager for Computer Vision at RealNetworks. “Secure, Accurate Facial Recognition (SAFR) is definitely AI-powered product that gives you next-generation visibility and situational interest for security professionals. ”
Unlike many other facial recognition platforms, SAFR was which is designed to accurately identify faces that will be in motion, under negative lighting conditions, partially obscured, or misaligned — real-life factors that cameras might capture from a natural setting.
SAFR also employs an integrated set of technologies to combine high performance with a good flexible architecture. This approach allows any forward-thinking company to support numerous practical use cases for enterprises and K-12 classes alike, from secure connection and gateway safety that will VIP loyalty and wedding venue monitoring.
Ultimately, RealNetworks is convinced of building the world’s the majority trusted facial recognition base through SAFR, leveraging ongoing improvements during both artificial intelligence and machine studying to continually raise the bar relating to accuracy and performance.
Using AI to build Sense of the Society Around Us
As much even as fear a technological takeover for sci-fi movie proportions, we should face the facts: Cameras have outnumbered humans on this subject earth. The current world citizenry is estimated at around 7. 7 billion, and there are something like 14 billion cameras at this present time. By 2022, that multitude is predicted to balloon to 45 billion. With nearly six cameras for every person on earth, many humans will never view almost all the visuals captured by they — there is simply too much data for the mental faculties to process.
“Instead, artificial intelligence will found yourself in the edge and make sense of all the visual information, ” Dan said. “With SAFR, we will enable customers in numerous verticals in the secureness space — home, commercial, enterprise, airports, hospitals, stadiums — to leverage our platform to be familiar with the world around him or her. ”
RealNetworks gives the software development kit (SDK) for SAFR designed to enable manufacturers of IoT devices during the edge computing space that will leverage its world-class facelift recognition technology through are located video. This will empower, for example, the head of security on a hospital or stadium that will derive new capabilities and insight from 100s of live video feeds in a way that was humanly impossible in earlier times.
“Device manufacturers can license our technology to bring truly world-class computer idea and facial recognition into the edge, ” Dan says. “We have SDKs designed to run on devices which include video conferencing cameras, home security cameras, mobile devices, wine cellar coolers, drones, cars, etc. ”
A very high Level of Accuracy together with Low Propensity for Disposition
Dan told us the fact that RealNetworks has observed striking advances in deep figuring out with convolutional neural networks (a types of artificial neural network useful to analyze images) over the last few years, and has worked to stay one step ahead during the quickly evolving space.
When assessed with benchmark of Labeled Faces during the Wild (LFW), a database of photographs manufactured by the University of Massachusetts to review unconstrained facial recognition in real-world conditions, SAFR revealed 99. 87% accuracy. What’s more, RealNetworks continually submits its algorithms into the National Institute of Principles and Technology (NIST) meant for performance evaluations.
NIST’s test results with April 2019 ranked any SAFR algorithm as the fastest and most streamlined among algorithms for undomesticated images, with a false non-match rate (the rate in which an algorithm mistakenly identifies two images within the same individual as different) of fewer than 0. 025. In terms of template extraction — practise of creating a facelift signature — NIST ranked SAFR’s algorithm among the many top seven worldwide.
“We are convinced of creating a computer vision platform that could be truly excellent, exhibiting high levels for accuracy and low levels of bias relating to variances in skin shade and gender — most are things that we happen to be continuously improving upon, ” Dan said. “We believe this systems offers tremendous benefits but will have to meet a certain bar of excellence for folks to retain customer believe. ”
Increase School Safety without charge in K-12 Schools
RealNetworks introduced SAFR into the market in 2018 as the free download for K-12 schools in the nation and Canada looking to better security and convenience for trusted members within the community. After guest complete a registration process at a tablet with staff affirmation, the technology will match faces in real time to streamline the obtain and check-in process.
“As a parent or simply a staff member, you can register yourself during the lobby of the faculty, ” Dan said. “The next time you come to front side gate of the campus, in place of standing there and punishing a buzzer that someone may or will not answer, with or without comprehension of who is there, a camera will recognise you and automatically unlock the entranceway. ”
In the same way that your technology makes security-conscious campuses out there to trusted members within the community, it may also be configured to spot persons of concern — which include expelled students or old employees — from within and outside of the school. In this strategy, the platform adds real value for Us schools, which have been inundated with deadly violence in the last few decades.
“We assistance our customers, particularly during the security space, see what matters in the real world, ” Dan said. “We own begun with facial realization, but we’ve also a short time ago launched with person detection and definitely will continue to expand our vision offering to address the needs of some of our customers for visual intelligence during the spaces they care related to. ”
A Marathon Technique to Improving Accuracy
Dan said building tremendously accurate facial recognition software may be a marathon, not a run. In addition to using feedback from the learning and enterprise security smaller communities, RealNetworks strives to continuously reduce the accuracy and speed of its algorithms to be able to at the forefront within the industry.
“The reality is there presently exist many facial recognition products available, and that number grows up every quarter, but they vary dramatically relating to performance, ” he says.
Ultimately, the mission at SAFR and RealNetworks could be to build the world’s the majority trusted visual recognition base. “In order to suit that responsibility, we believe we should design and develop a truly excellent product, ” Dan said.