What is RealNetworks and what are the domains it deals in?
RealNetworks is US based NASDAQ listed company, founded in the year 1995 and based at Seattle US. RealNetworks has offices across globe and in India its operating as a 100% subsidiary. RealNetworks has been pioneers in the field of media streaming. They were the first ones who brought music and video to internet. Our flagship products like RealPlayer is still very popular in many parts of the world. RealNetworks also has presence in Telecom domain and providing SMS related service, and some value-added services like Ring Back Tones to tier one mobile carriers across the globe. Since last 3-4 years RealNetworks has been doing research in the areas of Machine Learning and Artificial intelligence and have launched some exciting use cases.
What is Artificial Intelligence and Machine Learning and how does RealNetworks contribute to its development?
Artificial intelligence is a broader area in which you have machine learning and deep learning are subsets. In layman thoughts AI is where large amount of data sets is used to train machines so that the can-do specific task better than humans. Various algorithms are used to train the machine performance. Segment of AI where multi-layered neural network, similar to human brain are used to train machines are studied under deep learning.
RealNetworks has been working in the areas of NLP (Natural Language Processing) which is used for text and voice recognition. We have developed carrier grade A2P SMS classification platform named Kontxt based on machine learning. We have also been working in the areas of deep learning for face detection, face recognition and face analytics, object detection etc.
What is SAFR and how does it contribute to the Advanced Facial Recognition feature?
Advanced Facial Recognition (AFR) is one of the best facial recognition platforms for live video. Considering RealNetworks has over two decades of experience in the field of media streaming, we have combined facial recognition technology taking feed from live video. We have accuracy of face recognition over 99.87%; with mask SAFR can recognize face over 95%. We have added features like gender, age and sentiment estimation in SAFR along with face recognition. SAFR in its latest versions also detects face with person’s body and can be used in use cases like person counting, social distancing without need to recognize the face.
How does the Machine Learning Algorithm contribute to SAFR’s performance contributing to the advancement of Facial Recognition?
Various algorithms are used as stack on the overall solution to achieve best accuracy various features required in the solution. SAFR has been developed keeping in mind that the solution should be affordable and uses less hardware to achieve accuracy in minimum time. SAFR algorithm size is only 96kb making it one of the smallest algorithms with an accuracy of 99.87% and speed of less than 100 milliseconds. This means SAFR need less HW and face is captured, and result of recognition is achieved in less than 100 Milliseconds.
What are the system requirements that can enhance the working of facial recognition by SAFR and how is its integration with Milestone enhance the performance of Facial Recognition?
For deep learning solutions, we have large amount of data required for processing. The GPUs (Graphical Processing Unit) are 1000 times better in such processing requirements than CPUs. SAFR which can work only any configuration of HW but for better and faster performance we need IP cameras capturing the events, and GPUs being used for processing of the data.
Milestone is one of the best VMS (video Management Software) in the world. And SAFR is integrated with Milestone to the core level with face overlays on Milestone VMS. SAFR performance with Milestone in seamless manner with all the features what we have in SAFR stand-alone platform.
How does SAFR’s facial recognition contribute to the education sector, healthcare sector, and the retail sector?
For education and health sectors, SAFR can be used for general surveillance in the premises, face based touch less access control systems and attendance system. In Retail apart from above, SAFR can be used for demographic analysis of customers which can help retail sector for various policy decisions like product placement and inventory management, etc.
On a concluding note, how do you think SAFR’s Facial Recognition algorithm will help India as a country and its development? What are the other sectors that can be penetrated for better usage of the Facial Recognition algorithm?
In a country like India where population is large, such AI based solutions can largely proactively for securing and crime control, crime investigation, missing person tracing etc. During the time of Covid, SAFR can be used for contact tracing, mask detection, social distancing etc. Nowadays face based biometric attendance systems are replacing the finger-based attendance systems. SAFR can be used with existing CCTV cameras for making such seam less attendance systems. Health care can be benefited to large extend for skin disease detection, early detection of diabetic glaucoma etc. In human resources, SAFR can be used for sentiment analysis of workers in large factories of shop floor, helping HR manages to take various policy decisions.