ChinAI #272: Long Quan and the early wave of Chinese Computer Vision researchers
Greetings from a world where…
it’s courageous to be the first people to eat crabs [吃螃蟹的人]
…***Can you help ChinAI reach 200 paid subscriptions as we pass the 20k subscriber mark? Write it off as a business expense! As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: An Ethnic Chinese History of Computer Vision (part 1)
Context: Today, AutoX is a leading Chinese autonomous vehicle company, which fields a large fleet of fully driverless robotaxis. Its founder, Jianxiong Xiao, is a former Princeton professor who is known by arguably the best nickname in AI (“Professor X”). Raised in Chaoshan (a region of Guangdong Province, China), Xiao attended Hong Kong University of Science and Technology (HKUST) for undergrad. Back then, he studied under a man by the name of Long Quan, one of the early Chinese pioneers of computer vision. This week’s feature translation (link to original Chinese) traces the roots of the Chinese computer vision ecosystem back to Long’s time as a student in France.
Key Passages: On Long Quan’s breakthrough paper about image matching, which was accepted in 1988 to the second ever International Conference on Computer Vision (ICCV).
After the rise of the deep learning wave, Chinese scientists gradually became the main participants of the three major computer vision conferences (ICCV, CVPR, ECCV), published a large number of scientific research results, and won major awards. But in the 1980s, there were very few Chinese faces at the ICCV conference until the French student Long Quan broke through this situation in 1988.
In 1988, Long Quan, a Chinese student studying for a doctorate in computer science in the joint laboratory of the National Polytechnic Institute of Lorraine (INPL), the National Centre for Scientific Research (CNRS) and the National Institute for Research in Digital Science and Technology (INRIA), wrote a paper on image matching of indoor scenes using geometric constraints, under the guidance of his mentor Roger Mohr.
The paper was accepted by the newly started ICCV and became the lab’s first paper accepted into a top vision conference, and it also marked the first appearance of Chinese scholars at this visual conference.
Long Quan’s path to France and INRIA was supported by a Ministry of Education program.
After growing up in Taiyuan, Shanxi Province, he attended Northern Jiaotong University (now Beijing Jiaotong University) before entering a Ministry of Education program to study abroad in France: “When Long Quan was doing his undergraduate capstone project, he worked under then director Yuan Baozong and Zhou Zhongyu, who had just returned from France. The latter oversaw Long Quan's capstone and gave him a lot of guidance.”
His brother Quan Quan was also a big influence. “Quan Quan was admitted to Tsinghua University with excellent grades in 1978. A few months later, he was sent by the Ministry of Education to study abroad in Belgium. He was probably the first person from Taiyuan to study abroad after China’s reform and opening up.”
At INRIA, Long Quan became a leader in 3D reconstruction, an important method that translates 2D images into 3D space. This connected him with a group of Chinese scholars who were pioneers in the computer vision field.
From the article: “In France, Long Quan met Ma Songde (former director of the Institute of Automation of the Chinese Academy of Sciences) and Zhengyou Zhang (current chief scientist of Tencent), who had studied in France earlier. The three of them were collectively known as the ‘The Three Chinese Musketeers of Computer Vision.’ ”
During a sabbatical year, Quan visited CMU where he became friends with Harry Shum (who later became managing director of Microsoft Research Asia), and he also met Songchun Zhu, now director of Beijing Institute for General Artificial Intelligence, in Boston.
In the next stage of his career, at HKUST, Quan built small teams that drew on these relationships with other Chinese scholars.
Following Hong Kong’s handover in 1997, Hong Kong universities invested heavily in recruiting students and researchers from mainland China. After sixteen years in France, Quan joined HKUST in 2001 and established the “VisGraph” computer vision lab.
For instance, here’s how Quan recruited two doctoral students who later worked with Professor X: “Tan Ping and Wang Jingdong came to Long Quan's group through Microsoft Research Asia (MSRA). As mentioned above, Long Quan and Shum met at CMU and later became close friends. When Long Quan returned to HKUST, Shum was at MSRA in Beijing. Tan Ping and Wang Jingdong both became doctoral students through Shum's recommendation. The bond between HKUST VisGraph Lab and MSRA also began from this point.”
We’ll continue the story in part II next week. For now, check out the HALF TRANSLATION: An Ethnic Chinese History of Computer Vision: Long Quan and His Students
ChinAI Links (Four to Forward)
Should-read: The Wudaokou Origins of China’s Large Models (ChinAI #232)
Caixian Chen, editor of this week’s feature article about Long Quan, previously co-authored this history of the development of Chinese large language models in the Wudaokou neighborhood (in Beijing). That issue also links to another deep history of Chinese University of Hong Kong as a cornerstone of China's computer vision landscape.
Should-read: Do China’s Venture Capital and Private Equity Firms Welcome State Investment?
Stanford’s Center on China’s Economy and Institutions does a great job at distilling academic articles:
The average fund management firm dislikes government-linked capital and is willing to give up nearly $70 million in potential investments to avoid it. Dislike is highest among the best performing private firms…The researchers conclude that demand for government capital in China differs across different types of firms. The dislike of government capital, particularly among private, high-growth firms, highlights the potential for misallocation in parts of the VCPE industry where state actors dominate. In addition, the results suggest China’s government investors may not be able to attract the best firms to pursue their economic or societal goals.
The source publication: Emanuele Colonnelli, Bo Li, and Ernest Liu (2024). “Investing with the Government: A Field Experiment in China.” Journal of Political Economy.
Should-read: Is Xenophobia on Chinese Social Media Teaching Real-World Hate?
For The New York Times, Li Yuan tackles a puzzle: China has the world’s most sophisticated system to censor the internet, but the Chinese internet is laden with hate speech and disinformation. Contains many great details, including Yuan’s interview with Duan Lian, a Weibo influencer who often writes insightful articles that attempt to combat misinformation.
Should-read: Unit X - How the Pentagon and Silicon Valley are Transforming the Future of War
Raj M. Shah and Chris Kirchhoff have published an exciting book (coming out July 9, available for pre-order now) that chronicles how the Defense Innovation Unit aims to bridge the relationship between Silicon Valley and the Pentagon when it comes to emerging technologies. I’m pumped to assign this to my students!
Thank you for reading and engaging.
These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Also! Listen to narrations of the ChinAI Newsletter in podcast format here.
Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99