One Night at CVPR 2017 : Exploring IJCV and the Asia Research Force

This year, it’s in Hawaii.

The Computer Vision and Pattern Recognition Conference (CVPR), one of the top three events of its kind, is making Honolulu an even busier attraction this year.

According to organizers, the conference received 2,620 valid submissions, accepted 783 papers, 215 long and short orals, 3 parallel tracks, attracted 127 sponsors with $859,000 in sponsorship and close to 5,000 attendees up from just over a thousand a few years ago. All of these were happening at the Hawaii Convention Center as we wrote this.

Monday night, after another intense day of academic talks, scholars had to make that difficult decision again: “Which reception should I attend?” Numerous hi-tech companies, including Google, Facebook, and Intel hosted receptions around the conference for recruiting and partner recognition purposes.

One special event was the IJCV Asia Night. The keywords of the night were ‘IJCV’ and ‘Asia’.

Computer Vision Research: Three Top Conferences, Two Top Journals and One arXiv Platform – Where to Publish?

 

In the field of Computer Vision, it is well-accepted that there are three top conferences: Conference on Computer Vision and Pattern Recognition (CVPR), International Conference on Computer Vision (ICCV), and European Conference on Computer Vision (ECCV), as well as two top tier journals: International Journal of Computer Vision (IJCV) and Transactions on Pattern Analysis and Machine Intelligence (TPAMI).

Computer Vision and Pattern Recognition (CVPR) is hosted in the United States, while the European Conference on Computer Vision (ECCV) is hosted in Europe. The International Conference on Computer Vision (ICCV) rotates across Europe, North America and Asia. CVPR is held annually. ECCV and ICCV are hosted biannually and in alternative years of each other.

The International Journal of Computer Vision (IJCV) is published by Springer and the Transactions on Pattern Analysis and Machine Intelligence (TPAMI) is published by the Institute of Electrical and Electronics Engineers (IEEE). Both journals have impact factors of more than eight, while most engineering journals usually have an impact factors of only one or two.

To learn more about Computer Vision research journals, we interviewed Jennifer Evans, the Editorial Director of Computer Science at Springer and Professor Xiao’ou Tang, Chair of the Department of Information Engineering at the Chinese University of Hong Kong. Professor Xiao’ou Tang was a Program Chair of ICCV 2009 and will be the General Chair of ICCV 2019. He currently serves as the Editor-in-Chief of the International Journal of Computer Vision. Professor Xiao’ou Tang is the first Chinese Editor-in-Chief of IJCV.

“The fields of computer science are different from most scientific disciplines. In other fields, journals are usually the major publishing venues while conferences are mostly for socializing and exchanging ideas. However, since the field of computer science is developing so fast, conference has become another major venue for publication,” Professor Xiao’ou Tang said.

According to Professor Xiao’ou Tang, the review process for a computer vision journal is more rigorous thus can take longer, while conferences allow authors to publish their ideas more quickly. “Journal publication is usually more robust,” added Evans.

Usually, conference publications are limited to a maximum of eight pages, while the journals accept longer submissions. “Many comprehensive research papers in this field can take around 20 pages or more,” noted Professor Xiao’ou Tang. Journals allow the researchers to publish their work in more detail. Researchers can publish the “short” version at a conference first, and submit an extended version to IJCV later when the work is more mature.

Digitalization also plays an important role in computer science publications, especially the fields of artificial intelligence. “In the past, most university libraries would subscribe to journals and make them easy to access and we had to attend a conference to get access to the papers published in conferences. Nowadays we can access the conference papers online. arXiv has become a very important place for sharing our work with other researchers,” said Professor Xiao’ou Tang, “each channel has its own advantages and drawbacks.”

Professor Xiao’ou Tang patiently explained his duties as the current Editor-in-Chief of IJCV.

First, the Editor-in-Chief will look at the paper and assign it to an associate editor who is experienced in the field it covers. Then the associate editor will invite at least three reviewers to review the article and provide feedback. After taking their comments into account, the associate editor will decide if author of the submission will be asked for a revision. “There are always revisions and sometimes the process can take several rounds,” said Evans. Lastly, the associate editor will make a recommendation to the Editor-in-Chief to reject or accept the paper. The Editor-in-Chief then makes the final decision on the paper based on the recommendation of the associate editor and the reviewers’ comments.

Another important duty of the Editor-in-Chief is to appoint associate editors to join the editorial board. Keeping a high quality pool of associate editors is the key to uphold the quality of the journal. In the computer vision community, it is considered highly prestigious to be on the editorial board of IJCV and TPAMI. The number of associate editors from Asia is much smaller in proportion to the number of Asian researchers in the community.

In the past three years, IJCV has used more than 1,300 unique reviewers and 50% of them have reviewed more than one paper. “The Editor-in-Chief must be carefully selected. Only researchers who are well respected in their fields can make these decisions,” Evans stressed. As a former program chair of ICCV, Professor Xiao’ou Tang suggested that the “double blind review mechanism” of a conference is the key that makes publishing at one a fair game.
Compared to the traditional channels of journals and conferences, the emerging digital platform arXiv is the least restricted and most dynamic. Valuable work can be easily accessed and cited through arXiv. For these reasons, arXiv has become very popular. There could be more than 20 papers published on arXiv in the same day. This leads to the problem of filtering – the valuable work can also easily get buried. “It could be difficult for a not-so-famous researcher to get the first hundred citations if he solely published on arXiv, even if the work is valuable,” Professor Xiao’ou Tang noted.

 

The Rise of Young Asian Researchers: Look Back, Look Ahead

Prof. Kyoung Mu Lee is the General Chair of ICCV 2019. He currently serves as Associate Editor in-Chief of TPAMI, Area Editor of Computer Vision an Image Understanding (CVIU), Associate Editor of Signal Processing Letters (SPL), Machine Visions and Application (MVA), and IPSJ Transactions on Computer Visions and Applications. In his keynote during the IJCV reception, he talked about the AI boom in South Korea where many startups are emerging and traditional companies are turning to AI technologies. Meanwhile, academic activities, including KCVS and KCCV, are growing quickly.
Prof. Dahua Lin, came back to China and joined the Chinese University of Hong Kong after getting his Ph.D. at Massachusetts Institute of Technology. Dr. Lin commented on the latest national strategic issued by Chinese government in July. He believes that with the government’s supports on funding, data, talents and intellectual property, the nation’s AI research will definitely go further.
Chief Scientist of AI/ML at Google Cloud and the Director of the Stanford Artificial Intelligence Lab and the Stanford Vision Lab Professor Fei-Fei Li also gave an inspiring keynote during the reception. In her keynote speech, Li mentioned that there are two major trends in the world – one is the fourth industrial revolution led by intelligent machines while the other is the shuffling of geo-politics with the rise of the Asia. She mentioned that around one half of the first CVPR 2017 authors are of Asian descent.
Does that mean people of Asian descent are taking over the field?

“If you think of doing a research project like producing a movie,” Professor Xiao’ou Tang uses an interesting metaphor, “the first couple of authors are often like actors. The producers and directors are their professors or senior advisors. That distribution has not changed much. The number of high quality producers and directors is still small in Asia. It is too early to claim we own the field. We do not.”

To move forward in their research careers, people of Asian descent are also facing some culture specific challenges.
Prof. Ying Wu is an IEEE fellow who has been recognized for his ‘fundamental contributions to visual motion analysis and visual pattern discovery in computer vision.’ In his talk at the IJCV event during CVPR 2017, Prof. Ying Wu discussed the “name blindness” problem for Asian researchers.

Asian surnames appear first, whereas in many Western languages, surnames appear last. This leads to difficulties recognizing correct names in references. Many Asian authors of different works have the same surname. For example, the surname ‘He’ is shared by many people. It is important that the author’s full name is used when listed in reference. The night’s host, Prof. Xiao’ou Tang agreed. He joked that he has changed his first name to “XO”, similar to a prestigious and romantic Cognac Brandy. He joked that he was disappointed that some unfriendly friends have insisted that ‘XO’ is more recognized as a brand of Chinese Sauce, ‘XO酱’. Also, it symbolizes kiss and hug.

In addition to making citation easier for people with Asian names, many of the speakers noted that researchers of Asian descent could do more to contribute to the community.

Many scholars mentioned their appreciation to Professor Thomas Huang and Professor Takeo Kanade for being great role models for supporting this new generation of researchers. Professor Xiao’ou Tang, the host of the night, expressed that the spirit of the elder generation has inspired the current generation and he hoped they would also pass it along.

What else should this new generation of Asian researchers be thinking about?

In her speech, Professor Fei-Fei Li quoted one of her favorite contemporary philosophers, Shannon Vallor: “There are no machine values, machine values are in human values.”

Professor Fei-Fei Li then left the question open to everyone:

“AI will change the world. How will you change AI?”

 

Summary of IJCV Asia Night Keynotes

 

The night’s host Professor Xiao’ou Tang opened the night humorously by joking at the reception’s main entrée – Lobster. He recalled the lobster was quite affordable at the time when he was pursuing his Ph.D. at MIT. Now the lobster has become very expensive in China, to the point that even “small lobster” (小龙虾) costs as much as a Boston lobster in the old days. Thus he decided to invite the guests to enjoy lobster in Hawaii as it is more affordable here. The scholars all laughed and the night has started full of joy.

Then Professor Xiao’ou Tang introduced the openning keynote speaker – Dr. Shum, Executive Vice President heading Microsoft’s Artificial Intelligence and Research Group and an IEEE and ACM Fellow and Member of American Academy of Engineering.

Dr. Shum was asked about why some excellent scholars had moved from Microsoft Research Asia to other institutes during the Q&A time of his plenary speech at CVPR main conference. The night’s host, Professor Xiao’ou Tang introduces Dr. Shum with his own humorous answer to that difficult question – “It is because MSR trained everyone! The flow could only go one direction from MSR to other institutes.”

Dr. Harry Shum:

In his speech, Dr. Shum looked back at his time at MSRA and mentioned many of his colleagues. When Dr. Shum became the director of MSRA in 2005, Prof. Xiao’ou Tang took over the computer vision group from Dr. Shum. Many leading researchers have grown up from the group. Dr. Shum also expresses his appreciation and respect to Professor Tom Huang humorously. He jokes that Professor Huang has also taught him an “important” lesson aside from research: “After your kids grow up, you can travel with your wife!” Dr. Shum then expresses his wish to the new generation: “The current generation all look up to the previous generation. The current generation work hard so that the next generation would also respect and learn from the current generation; in this way, the community keeps growing and prosperous.”
Researchers and alumnus from MSRA are legends in this field, most of them have made important contributions and led academic trends. Many of the Asian people involved in computer vision research have become a force that can’t be neglected .

ICCV was first brought to Japan in 1990 and will be held in Korea in 2019. In 2005, ICCV was brought to China by Dr. Shum and many other great scholars. The tradition of Asian scientists at CVPR dates to 1983, when Takeo Kanade was the founding Chair of CVPR. Nowadays, a lot of Chinese companies have jumped into the AI game. It is a great time for computer vision. Every year, the number of conference attendees is skyrocketing, increasing collaboration between different Asian computer vision communities.

Dr. Shum said that he likes the old time ‘Enjoy doing CV and keep staying young.’ He encouraged young researchers to step up, successful researchers to help the next generation and give back to the community. At last, he called to unite the community, as united we stand stronger.

Prof. Katsushi Ikeuchi

Prof. Ikeuchi is the former Editor-in-Chief of IJCV. He is also a Principal Researcher at Microsoft Research Asia and an Emeritus Professor at the University of Tokyo.

Prof. Ikeuchi discussed the current status of IJCV. Its impact factors reached 8.22, extremely high in engineering. It takes an average 89 days from submission to the first decision and 23 days from acceptance to publication online. Moreover, Asian researchers play an important role both in submitting papers and downloading papers, making up 50% of this activity.
Prof. Ikeuchi encouraged everybody to support IJCV, since it plays a key role in the Asian community, and also Asian Conference on Computer Vision (ACCV), which is held biannually. He also proposed developing Augmented Intelligence instead of Artificial Intelligence.

 

Prof. Takeo Kanade

Prof. Kanade is the U.A. and Helen Whitaker University Professor of Robotics and Computer Science at Carnegie Mellon University. He is one of the world’s foremost researchers in computer vision, and serves as the Founding Editor-in-Chief of IJCV and the Founding Chair of CVPR.

Prof. Kanade accredited many pioneers in computer vision: Prof. King-Sun Fu, Prof. Thomas Huang, and others. Before Pattern Recognition and Image Processing, Image Understanding was the mainstream in computer vision.

At that time, Professor Kanade founded CVPR and IJCV together with several other respectful scholars. He recalled the story of how they made IJCV a top journal in computer vision by inviting all most valuable conference publications and set the submission standard extremely strict. Finally, he called for more talented researchers to join computer vision.
Dr. Rama Chellappa

Dr. Chellappa is the former Editor-in-Chief of TPAMI and the General Chair of CVPR 2017. He is a Professor of Electrical Engineering and an Affiliate Professor of Computer Science at the University of Maryland.

He credited his teachers, advisers and inspirers in his speech. Like other researchers, the successful scholar received tremendous help from the community. Dr. Chellappa appreciated many world-class masters, such as Prof. R.L. Kashyap, Prof. Azriel Rosenfeld, Prof. King Sun Fu, Prof. Thomas Huang, and Prof. Takeo Kanade.
Prof. Kyoung Mu Lee

Prof. Kyoung Mu Lee is the General Chair of ICCV 2019 and he currently serves as Associate Editor in-Chief of TPAMI, Area Editor of the CVIU and Associate Editor of SPL, MVA, and IPSJ.

Association for Computing Machinery (ACM Multimedia) 2018 will be held in Seoul, Korea from Oct. 26 – 28, 2018 and Pror. Kyuong Mu Lee is the Chair of the organizing team. Discussing his role at TPAMI, he mentioned that IJCV and TPAMI’s impact factors are both more than eight, and TPAMI’s is at all-time-high, up from 6.77 in 2015.
Prof. Fei-Fei Li

Prof. Fei-Fei Li is a Professor at the Computer Science Department at Stanford University. She is also the Chief Scientist of AI/ML at Google Cloud and the Director of the Stanford Artificial Intelligence Lab and the Stanford Vision Lab.

This is a historical time for AI and humanity, especially for Asian researchers. As we all know, Prof. Fei-Fei Li is one of the most famous Asian researchers in the CV community. Although Prof. Li has been involved in numerous influential works, such as ImageNet and the Visual Recognition Challenge, she’d rather humbly describe herself as an ordinary researcher who stands on the shoulders of giants. She said that many people of Asian background have made a huge impact on her career and shared some of her colleagues and friends’ special moments when she was a young professor including her “Academic Grandpa” Jitendra Malik in UCB, Prof. Song-Chun Zhu in UCLA, Dr. Hurry Shum in Microsoft, Prof. Kai Li in Princeton. She also acknowledged two students: Jia Deng, who was the student collaborator of ImageNet, and Jia Li, her former student and now colleague at Google, their lifelong collaboration and friendship influences her work everyday.

“With great power comes great responsibility,” Li quotes Spiderman to illustrate the power and responsibility of this generation of researchers. There are two major themes in the world at this time:

1. We’re going through the fourth industrial revolution led by intelligent machines and a fast-growing cyber community;
2. A shuffling of geo-politics, or “the rise of Asia,” The CEOs of many technology companies, authors of CVPR, and most importantly, China, the first country realizing the power of AI and to release a government white paper on AI.

However, this could also be a testing time for humanity. Take chat-bots: they could help remote people access healthcare and education, or become a source of cyber-bulling and disinformation. Another example is self-driving cars. They may prevent us from traffic accidents and help keep the environment healthy, but we should also consider the impact of automation on the labor market. Tech giants and startups are pursuing the power of AI but will it help to build a better life? Will AI help the rich and powerful to become richer and more powerful, creating an even more polarized world? History has shown humanity has an innate desire to innovate which could also lead to catastrophe.

At the end of her keynote, she quoted one of her favorite contemporary philosophers, Shannon Vallor: “There are no machine values, machine values are in human values.”
Dr. Zhengyou Zhang

Dr. Zhengyou Zhang is the General Chair of CVPR 2017. He is a Principal Researcher and Research Manager at Microsoft Research. He is also an IEEE and ACM Fellow.

Like all other speakers, Dr. Zhang first gave his thanks to his advisers and mentors. His first paper on CV was accepted at the second ICCV in 1988. At that time, his paper was poorly written but the community was more tolerant. He then improved his writing and became a book author.

Dr. Zhang gave his advice to the young generation: “Be calm and stay focused, take your time to pursue work that is meaningful.” He said his own productivity cycle was one paper every two years. He encouraged the young generation not to rush – “If one wants to make big, useful work, focus is key. Once your works reach a certain level, others will benefit.”
Prof. Ying Wu

Prof. Wu is a Program Chair of CVPR 2017. He is currently Professor at the Department of Electrical Engineering and Computer Science at Northwestern University.

Prof. Wu is an IEEE fellow for his fundamental contributions to visual motion analysis and visual pattern discovery in computer vision. The highly-respected researcher views himself as a service worker to the community, especially the Asian one.
Prof. Yanxi Liu

Prof. Liu is a Program Chair of CVPR 2017. She is Professor at the Computer Science Engineering and Electrical Engineering Department of Pennsylvania State University.

Prof. Liu gave a speech about what a pattern is. There are many ways to define a pattern, such as motion, texture, and symmetry. The fundamental property of a pattern is that it is recurring. We can find patterns through unsupervised searches, find semantic meaning, or clustered subspace.

During the speech, she showed a video about puffer fish creating amazing and perfect circles. Patterns exist everywhere, not just in human society. Symmetry is something we should pay attention to.
Sponsors

During the banquet period, three sponsors, Microsoft Research, SenseTime, and China Unicom, gave their own short speeches that introduced their businesses and attitudes towards AI.

 

---

Authors: CZ, QW | Editor: Nicholas Richards