Computer Vision: Bringing Sci-Fi to Reality

In May 2016, the recently returned series “Person of Interest” had already depicted a series of sci-fi scenarios where crimes are monitored, analyzed, and recognized by machines back in 2011. A mysterious billionaire invented an artificial intelligence system with cognitive abilities, using cameras as eyes to identify criminals and predict their actions. This system not only detects criminal activities but also helps ordinary people who are in harm’s way — it can guide machines to rescue victims in advance.

Five years later, with the significant progress in AI across various fields including voice recognition, computer vision, natural language understanding, and healthcare, and surpassing human capabilities in areas like speech and facial recognition, those seemingly distant and sci-fi scenes from television are gradually becoming real in people’s lives.

Computer vision is a new field that brings sci-fi into reality.

The process described above, which involves collecting facial images through cameras and identifying individuals based on geometric features and relative positions, is what is commonly known as facial recognition. In China, facial recognition has already begun to be applied in security, finance, and healthcare, with increasingly sci-fi scenarios becoming real due to the widespread application of computer vision.

Globally, there are three top research institutions in the field of computer vision — MIT, UC Berkeley, and the Oxford University lab in Europe. Staff from these institutions are constantly being recruited by tech giants like Google or starting their own ventures, becoming a crucial part of the entire technological development process.

The development of computer vision in China began in the late 1990s. With the rapid economic growth and continuous improvement in education levels, many industry elites who once went abroad are increasingly returning home, rapidly accumulating high-quality talent in China, which provides a human resource foundation for the rapid development of technology.

At the same time, with the advent of the computer age, the rapid advancement of networks, improved big data computing capabilities, and the seamless deployment of cameras have laid a solid foundation for the software and hardware technologies of computer vision in China. For more than a decade, the applications seen in many sci-fi films have become achievable through computer vision — “seeing” people’s every move and “predicting” their future behavior.

Today, with a solid talent base and excellent software and hardware technologies, the world of fantasy is being realized one by one. Driven by numerous startups, China’s development level in the field of machine vision is now on par with the world’s most advanced countries like the USA and Israel.

Computer Vision

The Overlooked Pearl of Artificial Intelligence

In the field of computer vision, technology has long surpassed the image recognition accuracy of the human eye and is widely applied in public security, finance, and information security. However, these significant social values achieved have not garnered public attention; people are more excited about AlphaGo’s victories than the actual applications of technology in life. This is because people intuitively believe that recognizing images is an easy task, while defeating the world’s top chess players is a greater challenge.

In fact, from the perspective of AI development, Go is a game with logical rules and measurable calculations, where the difficulty for the human brain lies in the vast amount of calculations and the judgment of the macro situation on the board; while image recognition presents a broader randomness and uncertainty in the information capture and logical analysis layers. Classifying and parsing information from images through machine learning to ultimately extract valuable structured data is an extremely challenging research topic, with the transition from academia to industry taking decades.

Compared to other advancements in computer vision, natural language understanding, and speech recognition, AlphaGo certainly has its epoch-making significance: it not only shortens the intelligence gap between machines and humans but also poses a more frightening prospect — in the future, the perceived intelligence gap between people will no longer be an irreparable innate difference but rather an ability acquired through tools. This will overturn the self-evaluation of human value.

The technological advancements in computer vision will bring about far greater transformations in real life than academic breakthroughs and philosophical reflections; technology is helping to expand the boundaries of human capabilities.

Technological breakthroughs in the field of computer vision were first applied in vehicle recognition. For instance, Yitu’s vehicle recognition system based on video images helped the police solve a burglary case worth over 100,000 yuan when it was first launched in Suzhou. After the suspect robbed the place, he drove away from the neighborhood, and the police quickly identified the vehicle using Yitu’s “vehicle recognition system” through brand filtering, solving the case in just ten minutes.

Compared to vehicle and object recognition, facial recognition has a wider range of applications and scenarios, currently capable of performing static facial comparison recognition in portrait libraries of over a billion images, and has been successfully applied in security systems for the Youth Olympic Games and the Zhuhai Airshow.

The scene in Xiamen is strikingly similar to the plot in “Mission: Impossible – Ghost Protocol” where Hanaway uses a facial recognition system at a train station to quickly apprehend a suspect. And today, it is happening in reality.

In addition, besides the public security sector, vehicle and facial recognition technologies are also widely applied in finance and other fields. For instance, China Merchants Bank was the first in the country to deploy facial recognition technology on VTM visual counter machines across 1,500 branches for tasks requiring on-site identity verification, achieving a 98% accuracy rate during testing, with an error rate of only one in a hundred thousand, while the human eye’s error rate is one percent. This means that the accuracy of machines has far surpassed that of the human eye.

During my postdoctoral studies at the MIT Artificial Intelligence Lab, I studied under Allen Yuille, one of the pioneers in computer vision. He introduced me to the world’s top computer vision technology circle.

With a simple curiosity about artificial intelligence, I returned to China and co-founded Yitu with my high school classmate Lin Chenxi, who was then responsible for the Aliyun Flying Sky project, to begin technological exploration in the field of machine vision. I am more skilled in computer vision, while Chenxi excels in cloud computing and big data technologies. He led the development of the largest domestically owned distributed cloud operating system, which gives us a solid reserve and advantage in big data technology.

It is precisely due to the accumulation and practice in the technology field that Yitu had the opportunity to partner with Huawei, a highly respected company and a banner for China’s advancement into the world. In the construction of safe cities, Yitu can effectively combine machine vision technology with Huawei’s over 20 years of experience in the ICT field; at the same time, Huawei’s open ICT platform and ecosystem can help Yitu quickly form reliable solutions to serve users.

For a startup like Yitu, Huawei enables us to see the world from a giant’s shoulders. In the future, we have the opportunity to leverage Huawei’s global network to serve users in more countries and build safer cities.