Winner of AI Heroes Pioneer Award: Practitioner Yan Shuicheng

category:Internet click:511
 Winner of AI Heroes Pioneer Award: Practitioner Yan Shuicheng


Winner of AI Heroes Pioneer Award: Yan Shuicheng (Source: Netease Technology Channel)

Wen/Ding Guangsheng

Professor Yan Shuichengs office is small and full of various engineering prototypes.

I came to the office one step at a time. The intelligent doorbell at the door had found my visit. Professor Yan Shuicheng came in a hurry. This is an intelligent hardware product of 360. It is supported by Yan Shuichengs 360 Artificial Intelligence Research Institute. At present, 360 sells dozens of intelligent hardware products, which is also Yan Shuichengs original intention and mission to 360.

Love without marriage is hooliganism. He uses this humorous metaphor to describe the importance of AI landing practice. He believes that the academic research on AI is like love, while the industrial research on AI is more like married life.

Over the past three years, Yan Shuichengs 360 Institute of Artificial Intelligence has been working in four directions: vision, voice, semantics and big data, providing technical support to business departments, focusing on four AI engines: motion, interaction, vision and decision-making.

Over the past three years, Yan Shuicheng, as an AI practitioner across academia and industry, how to understand their love, hatred, hatred and how to become a master step by step? In his own words, Our generation was lucky. Now many well-known scientists in the field of computer vision were born around 76 years ago. Li Feifei, Wu Enda, Sun Jian and so on.

Whats beyond luck?

Academia Pursues Limits

Yan Shuicheng spent nine years in the School of Mathematics Sciences, Peking University. He studied Computational Mathematics and Applied Mathematics. Although he was in the Department of Mathematics, he followed his tutor, Professor Cheng Qiansheng, to study subjects related to pattern recognition and machine learning.

In 2001, Li Ziqing, an expert in face recognition and intelligent video surveillance, joined Microsoft Asia Research Institute. He quit his tenure at Nanyang University in Singapore in 2000, joined Microsoft Asia Research Institute, and developed a face recognition system EyeCU at Microsoft.

Yan Shuicheng spent almost all his PHD time at Microsoft Asia Research Institute. Later, he was mainly taught by Dr. Zhang Hongjiang, a world-class scientist in the field of multimedia research. Zhang Hongjiang is known as the founder of the research field of computer video retrieval. He is also one of the founders of Microsoft Asia Research Institute.

Zhou Zhihua, Yan Shuicheng, Mei Tao and other active scientists were disciples of Zhang Hongjiang. The picture shows Zhang Hongjiang.

Microsoft has a very good atmosphere, very free at the same time, there are many directions, so that I was in computer vision and machine learning in both directions have a very wide range of involvement. During this period, he Xiaofei (former Dean of Droplet Research Institute, now CEO of Flying Technologies) jointly linked Subspace Learning and Manifold Learning in a clever and simple series. They later published four single papers citing over 1,000 papers, which had a wide impact in this field.

But when he had to make a choice, if he chose to keep Microsoft, Yan Shuicheng felt that he could see three or five years later at a glance, or should go abroad. He first came to Professor Tang Xiaoous laboratory at the Chinese University of Hong Kong to continue his research on face recognition. Professor Tang Xiaoou traveled back and forth between Beijing and Hong Kong at that time. During this period, Yanshui achievements led us to promote project research. His leadership ability was trained and exercised, which laid the foundation for the future.

More than a year later, Yan Shuicheng went on to the University of Illinois at Champaign (UIUC), where he was taught by Professor Thomas Huang, an academician of the American Academy of Engineering and the founder of computer vision. He met a group of students with excellent algorithms and programming skills, including Zhou Xi, Han Xu and Jianchao Yang, who were later known to all. Together, the group contributed 7 of 9 CVPR articles, and the remaining 2 were quick to get the perfect experience of other meetings.

The dust settled down after a circle.

Yan Shuicheng returned to Asia and entered the National University of Singapore, opening up his most productive eight-year academic career. He believes that there are advantages over both the United States and China. A professor in the United States needs to spend a lot of time writing project applications, and the hit rate is very low. In China, an energetic professor may spend too much time dealing with things other than academia to maximize it. Use your precious time.

In Singapore, Yan Shuicheng wrote a project application about one year. Eight years in Singapore has been my most academic focus for eight years, and the output is very large. His team has been CVPR and ACMM two of the most conference papers in the world on many occasions. His team was the first Asian team to compete in the Computational Vision Core Competition PASCALVOC and won the championship in object recognition for three consecutive years from 2010 to 2012. At the same time, the team also participated in the ImageNet competition many times and won the championship. The 1x1 convolution of the network-in-network model proposed during the period became the basic module of almost all subsequent computer vision CNN variant models.

At present, he has published about 600 high-quality academic papers in international high-quality journals and conferences. His papers have been cited 40,000 times, H-index 92. He is currently the highest H-index scholar in ComputerScience & Engineering field in Singapore. He has been selected as TRHighly-cited searchers four times in 2014, 2015, 2016 and 2018.

National University of Singapore

In Yan Shuichengs view, in academia, it is the pursuit of the ultimate, more is the vision of technology and products, or continuous technological sublimation, most of the time do not really need to land the product.

In this way, Yan Shuicheng germinated the idea of joining the industry.

Industry is in business.

Later stories must be well known. When Yan Shuicheng came to 360 company to communicate with 360, he met 360 immediately. On the one hand, Yan Shuicheng was a technological master who had studied deeply for more than ten years. On the other hand, he was eager to build AI research institute. In 2015, Yan Shuicheng chose to go north and began to devote himself to industrial research.

In the industry, Yan Shuicheng began to be a businessman. He returned many years of technological accumulation to business, and the landing of products became the core of his work. He believes that the most important thing at this time is the shift from academic thinking to industrial thinking.

He humorously compares this transformation to two people falling in love, two people falling in love is a compulsory course, but married life is the essence, love without marriage as a purpose is hooliganism, and in the industry must take product landing as its guiding ideology.

What is unique is that Yan Shuicheng is also an associate professor at the National University of Singapore. His research and development work is more relaxed because of his cross-disciplinary background. In 2016, Human Segmentation showed commercial possibilities on mobile phones. A European research team began to show that Yan Shuichengs team was interested but had no choice but to charge too much for each other. Our NUS laboratory, thanks to the accumulation of previous technology, took only one month to achieve success, and began to apply human body segmentation technology on mobile phones.

He combines academic research with industrial practices and sparks a spark.

He also suggested that professors in academia should embrace the needs of industry more when choosing topics and really aim at social values. For students, most of them will eventually enter the industry. Really useful topics are also beneficial to studentslong-term development. Its hard to produce value for a superb demo, but he also admits that people need to explore the boundaries of Science in this era, and many laboratories often find surprising things.

For AI experts such as Li Feifei, Li Jia and Zhang Tong returning to academia, Yan Shuicheng mentioned that the difference between academia and industry is obvious to all. The key is whether they can fit in with their own values and pursuits. If there are deviations, it is inevitable to choose to leave. However, he believes that the ideas of these scientists have penetrated into their enterprises and will certainly have a positive effect in the future.

Please answer 2019

The AI industry will move in a more rational and vertical direction.

Yan Shuicheng frankly said that in the hot autopilot track last year, we were generally facing L4/L5. This year we can clearly perceive that we have returned to rationality, more people choose to focus on limited scenes, ports, highways, parks and so on.

On the other hand, computer vision companies will become more focused on scenarios and become a whole solution provider, not just a pure provider that only provides SDK output or technology, and we need to think about it in the long run.

More rational and more focused is the summary of Yanshuis industry trend. In 2016, as long as it is an AI company, people will invest crazily. In 2017, you will see if the company has landing potential. But in 2018 and 2019, people begin to really value the ability of an AI company to make money.

From the technical dimension, Yan Shuicheng expects the real outbreak of AI chips and 5G in 2019. In the past two years, we have only tried and explored. In 2019, we hope to see the rapid development of AI chips and 5G products.

Evolution of Mobile Communication Network Technology

We have noticed that the year 2019 will be the first year of 5G applications, which will gradually reach industry consensus. This year, more than 30 5G mobile phones will be launched worldwide. As the fifth generation mobile communication network, 5G network has a peak theoretical transmission speed of tens of gigabytes per second, which is hundreds of times faster than the transmission speed of 4G network. The whole ultra-high-quality film can be downloaded within one second. With the birth of 5G technology, it will be used. The era of smart terminals sharing 3D movies, games and super high-quality programs has come to us.

Yan Shuicheng pointed out that the large-scale application of AI chips will make the computer vision industry usher in an opportunity again. With the support of AI chips and 5G, he is still optimistic about the development of AI industry in spite of the challenges in the cold winter of capital. We have invested so many years, we should consider commercial returns.

However, for those slow-moving AI projects, the next year will be quite tough.

Postscript:

An interesting point of view

Now many well-known scientists in the field of computer vision were born around 76 years ago, Li Feifei, Wu Enda, Sun Jian and so on.

This reminds the author of Malcolm Gladwells remark in Aliens that Bill Gates and Steve Jobs were born in 1955, around 1955, during the period of the computer revolution. If you were born too early, you could not own a personal computer. If you were born too late, the good chance of the computer revolution was taken up by others, and the fantastic people, why they are Magic, thanks for the opportunity.

A commentary

Netease 2018 China AI Heroes Wind and Cloud List Business Innovation Pioneer Award winners are Zhou Xi and Zhu Long, I very much look forward to these two young scientists, they can cast aside the academic aura, step by step in the business sector to move forward, their company may not be the fastest growing, but the future will certainly achieve extraordinary achievements.

Source: Netease Intelligent Responsible Editor: Ding Guangsheng_NT1941