RealNetworks wants to dig into the billions of markets
What industry is the hottest today? The answer is undoubtedly artificial intelligence.
From technology giants to large Internet companies to startups, they have flooded into this market. There are Microsoft, Google, Amazon, and RealNetworks in foreign countries, and domestic technology companies such as Tencent, Baidu, Shangtang, and contempt.
Data from iResearch indicate that the global artificial intelligence market size will be about 119 billion yuan in 2020. In the next 10 years, artificial intelligence will evolve into a $ 200 billion market with huge space. According to the China Artificial Intelligence Development Report 2018 released by Tsinghua University, as of June 2018, there were 2028 artificial intelligence companies in the United States, ranking first in the world, and China ranked second with 1011 artificial intelligence companies.
Breakthrough point of artificial intelligence-face recognition
Technically, artificial intelligence can be divided into computer vision, speech recognition, natural language understanding, and machine learning and knowledge maps. In computer vision, face recognition has been the first to be applied, becoming the first burst point.
According to relevant analysis reports, the global face recognition market was US $ 3.85 billion in 2017 and is expected to reach nearly US $ 10 billion by 2023, and the overall market size will maintain a growth rate of about 20%.
In China, there are three major Internet companies-BAT, and four computer vision unicorns-Shangtang, Deaf, Yuncong, Yitu, and other technology companies. Now, this competitive and vast market welcomes a new entrant: RealNetworks.
Recently, the company introduced its well-known face recognition platform SAFR into China and signed a contract with a domestic company. In response, reporters interviewed Mr. Reza Rassool, Chief Technology Officer (CTO) of RealNetworks.
As a world-renowned provider of audio and video solutions, RealNetworks was founded in 1995. That year, RealNetworks launched RealAudio, the first Internet audio streaming solution. Since then, the company has begun to develop rapidly, introducing more and more products and services, including RealMediaHD, RealAudio, RealPlayer, Gamehouse, mobile services, Realmedia, KONTXT, and so on.
According to Reza Rassool, he is from the UK and has lived in the United States for 25 years. I have 24 patents in the United States and 40 patents worldwide, and I am a true technical expert. In the 1990s, he joined a team that launched the world's first movie editing software to run on a computer. Later, he also participated in the development of a video-on-demand system and researched video and audio processing technology, technology for cochlear implantation and biological vision.
In 2000, they founded a company called Widevine, which is committed to providing online video payment security technology, and was later acquired by Google.
Talking about why go to RealNetworks? Reza Rassool said, I have been a loyal user of RealNetworks throughout my career, so when I have the opportunity to come to this company. Accepting this job is a natural choice.
As the CTO of RealNetworks, he is responsible for the technical management of the entire company and the development of two products: the digital codec RealMedia and the face recognition SAFR.
Today, artificial intelligence technology has gradually become a basic technology, which has become the two key issues that place equal importance on data in enterprises. For RealNetworks, SAFR is the company's fist product for artificial intelligence.
Reza Rassool said, As the inventor of streaming media, RealNetworks has been innovating. We hope that when decoding a video, we can use machine algorithms to understand what is in the video. This can be understood as the recognition, tracking and Monitoring. And face recognition is the first feature of computer vision.
Through such technology, RealNetworks can know, track and discover the content of the video, and based on this help customers develop more video features.
We are constantly exploring machine learning and artificial intelligence, hoping to bring more power to our products, he said.
Prior to interviewing Mr. Reza Rassool, RealNetworks released its financial results for the third quarter of fiscal 2018. In the financial report, the CEO of this company specifically mentioned the face recognition platform SAFR, which can be said to have high hopes for it, hoping that it will become the company's new growth engine.
Reza Rassool said, Machine learning has become a very successful computing method that can solve many problems that cannot be solved by the human brain. We take it as a strategy and hope to achieve it better in RealNetworks' future system.
It is understood that the SAFR face recognition platform began operation in 2016 and was officially released on July 17, 2018. After its release, SAFR has been certified by the National Institute of Standards and Technology (NIST) for accuracy, which is equivalent to getting a certificate.
According to him, SAFR's name comes from four words: Secure, Accurate, Facial, and Recognition.
At the interview site, Reza Rassool personally demonstrated SAFR's face recognition platform for reporters. Computer-based camera, after data collection, the computer displays his information, including gender, age and expression. And without data collection, you can't tell who you are.
Less processing and lower costs
Through computer detection, SAFR can send information to the cloud, request it from the cloud, and match who you are. In addition to cloud identification, SAFR also supports processing information locally. RealNetworks focuses on how to handle things efficiently, with less computing and lower costs, so that the technology can be better disseminated, he said.
Reza Rassool demonstrates SAFR platform to reporters
At present, RealNetworks has signed a partnership with Japan's very large system integrator Net One Partners and Brazil's Seventh Company to integrate SAFR face recognition into their products and systems.
Also, SAFR will be provided to K12 schools (from kindergarten to high school) in the United States and Canada for free this summer. In the United States, for example, there are often school shootings. We want to make schools in the United States and Canada more secure. He added. However, the camera is not installed in the classroom, but on and off the campus to monitor who enters the school and whether this person is eligible to enter the campus.
In fact, face recognition, as the first artificial intelligence application, involves many factors, such as security and privacy. Reza Rassool pointed out to the Tianji Network reporter: From the beginning of this project, the top level of RealNetworks has decided to make a socially responsible project.
From this, we can understand why SAFR was the first to get a landing at K12 schools in the United States and Canada.
He said, We hope that this facial recognition technology will bring more convenience to life and allow everyone to have a more efficient life with some information provided. Of course, this technology can also be applied to the business field To promote business development.
In fact, there are still many shortcomings in many face recognition platforms, such as the inability to recognize faces in motion, faces with occluders, and faces after makeup. He believes that the challenge of face recognition technology at this stage is: how to continuously improve the performance of face recognition systems under natural conditions, continue to reduce hardware costs, and increase the speed of face recognition.
Competition is core competitiveness
From technology to the market, everyone competes for core competitiveness. A lot of competitors in the market do not have the corresponding strength and technology to solve the above problems, and this is what RealNetworks is good at and continues to invest in. Reza Rassool said.
Compared to competitors, he mentioned three advantages of RealNetworks' face recognition platform SAFR.
First, SAFR can respond faster than competitors with equal accuracy. He said, In our NIST test, RealNetworks responded fastest by making an API request to the United States locally. With this technology, we can achieve real-time face tracking and monitoring.
Second, the trained model can be localized. According to him, the machine learning model used by SAFR is relatively small and flexible. It can be implemented locally and processed at the edge. Their competitors often use very large machine learning models, which are 20 times or even 30 times that of RealNetworks.
Third, SAFR's algorithm is very fair, and it can accurately identify regardless of the person's skin color, age, gender, and geographic location.
This involves a problem, how to ensure the fairness of the algorithm. Reza Rassool introduced the company's approach. RealNetworks guarantees the fairness of the algorithm from two aspects: one is to strictly manage and control the data; the other is to collect various data and photos worldwide, regardless of any region, any race, and All are photo data from real scenes, not specially modified photos or ID photos.
Although RealNetworks has its own unique advantages, how to compete with other face recognition companies.
Reza Rassool talked about the advantages of SAFR: Companies like Google and Microsoft generally use the path of accepting customer data, uploading it to the cloud, and then returning the result.
For example, Amazon will charge $ 1 for every 1,000 API requests. It is more economical and convenient for small equivalents, but for many large amounts of data, such as real-time video, there will be thousands of faces in 30 frames per second. Such requests will be very expensive.
RealNetworks has a technology that can be quickly promoted. Based on that users only use a small amount for each upload to the cloud, just upload 80x80 pixels to identify a face.
We hope to achieve cost-effective face recognition with real, real-time video. We believe that our products have advantages in the market, he said.
According to the China Artificial Intelligence Development Report 2018 released by Tsinghua University, the size of China's artificial intelligence market reached 23.7 billion yuan in 2017, a year-on-year increase of 67%. The market size of computer vision, language, and natural language processing accounted for 34.9%, 24.8%, and 21%, respectively, and the total market size of hardware and algorithms was less than 20%. It is expected that the growth rate of China's artificial intelligence market in 2018 will reach 75%.
In the field of face recognition, there are both large Internet companies in China, such as Ali and Tencent, as well as unicorns like Shangtang and Deaf. When the reporter asked if he was ready to enter China, Reza Rassool replied: For the new business, we have excellent engineers. As the company's chief technical officer, I can guarantee that we are fully prepared technically.
RealNetworks hopes to become the leader in this field in any of its businesses, and the same is true in face recognition. We are a 25-year-old global company that came to China and hopes to have a long-term presence in China. Business. He said.
In fact, RealNetworks just signed a strategic cooperation agreement with Shenzhen Hongtian Hi-Tech to bring its high-precision facial recognition platform SAFR to China. It is reported that Hongtian Hi-Tech has strong communication technology and product development, wireless Internet value-added service operations, electronic design and manufacturing capabilities, and has done a good job in digital broadcasting and television, mobile communications and broadband Internet across networks.
Reza Rassool said, Working with Hongtian, we mainly value their talent system and hope to work with them to develop products and jointly develop SAFR. Our initial cooperation intention is to apply such technology in retail, hotel industry and security. aspect.