Shulou(Shulou.com)06/02 Report--

2019-10-28 18:57:34

ICCV2019 officially opened in Seoul, South Korea this morning. A series of exciting activities such as Workshop, Tutorials and Doctoral Consortium with rich themes will be held one after another in these two days.

COCO is the highlight of ICCV 2019 and the most influential general object detection challenge in the field of artificial intelligence. It won twice in 2017 and 2018. According to the latest news from the front, this conference made unremitting efforts, won the first of the three items of COCO object detection (Detection), human key points (Keypoint) and panoramic segmentation (Panoptic), and achieved a remarkable record of three consecutive COCO championships. It is worth mentioning that the most innovative algorithm also won the newly established Best Paper Award (Best Paper Award) this year.

(picture from farsighted Technology Weibo)

In addition, Xiangshi also won the champion of Wider Challenge 2019 pedestrian detection and VOS 2019 video object segmentation. After 22 global championships in AI technical evaluation, Kuangshi won 5 championships in the ICCV 2019 academic Top Competition.

(picture from farsighted Technology Weibo)

Last Friday, we released 15 entries for each person. How much do you know about Chinese papers? In-depth analysis of the things behind the admission of ICCV2019 papers! This paper analyzes and interprets the admission data of ICCV papers.

There are friends in the background to leave feedback, their own school obviously sent N articles, why not on the list? Question our results.

In the above, we have made it clear that there are no full statistics, and the statistical results are somewhat different from the facts, so please forgive me for those units that do not have statistics.

Fortunately, all the papers have been published on ICCV2019's official website in the past two days, with a total of 1075 papers. So the little friends of AMiner quickly grabbed the data again in the past two days, and then we counted according to the institutions to which all the authors of each selected paper belonged, and in the case of an author with multiple units, according to the author's first unit statistics, to analyze the number of articles issued by each organization in ICCV2019.

ICCV2019 papers Collection address: http://openaccess.thecvf .com / ICCV2019.py

Analysis from the point of view of mechanism

Of the 1075 selected papers, more than 40 are selected by five major organizations, three of which are from Chinese academia.

Among the top 10 institutions, seven are from academia, four and three from Chinese and foreign scientific research universities, and three from industry.

(manual statistics of the above data, if there are any omissions, please leave a supplementary message.)

Among them, Google + DeepMind combination topped the list with 47 articles. According to the ranking of the number of contributions released earlier, Google ranked fourth with 100. In the end, 47 high schools reflect the absolute strength.

In second place was the Chinese Academy of Sciences, with a total of 45 selected articles.

Followed by Tsinghua University in third place, 44 articles were selected.

The Chinese University of Hong Kong and Facebook each had 42 articles selected, ranking fourth.

Peking University and the University of California ranked fifth and sixth with 32 and 31 articles respectively.

Twenty-nine articles from Oxford University were selected, ranking seventh.

Shangtang Science and Technology 28 articles, ranked eighth. It is also the only Chinese company in the top 10 list. Shangtang Technology, which has been founded for only five years, is a rising star. It can be said that there is unlimited potential to achieve such a brilliant result this time.

Stanford University and the Federal Institute of Technology Zurich each had 27 articles on the list, ranking 10th.

Among the more than 20 selected papers, 26 were from Carnegie Mellon University, 25 from Massachusetts Institute of Technology, 24 from Sydney University, 22 from Adobe, 21 from Nanyang University of Technology, 21 from Beijing University of Aeronautics and Astronautics, 21 from the Origin of artificial Intelligence, 20 from the University of Maryland and 20 from Microsoft.

A total of 28 papers have been selected from less than 20 to more than 10, including 9 from Chinese academia and 6 from Chinese industry.

Among them, there are 19 articles from Chinese Academy of Sciences, 18 from Sun Yat-sen University, 18 from Shanghai Jiaotong University, 15 from Dalian University of Technology, 14 from Huazhong University of Science and Technology, 13 from University of Science and Technology of China, 10 from Zhejiang University, 10 from Tianjin University and 10 from Northeast University.

There are 17 articles on Baidu, 15 articles on Tencent, 12 articles on Huawei, 12 articles on Alibaba, 11 articles on science and technology and 10 articles on JD.com.

In addition, IBM, Nvidia and Samsung also had 14, 13 and 11 papers respectively.

More than 6 papers were selected, including 8 from Xiamen University, 8 from South China University of Technology, 8 from Nankai University, 8 from Harbin Institute of Technology, 8 from Xi'an Jiaotong University, 8 from Hong Kong University of Science and Technology, 7 from Nanjing University, 7 from Wuhan University, 7 from Beijing Institute of Technology, 7 from Nanjing University of Science and Technology, 6 from Hong Kong University of Technology and 6 from Northwestern Polytechnic University. 6 articles of Beijing University of posts and Telecommunications.

Among the institutions not listed in the picture above, Fudan University (5), Xi'an University of Electronic Science and Technology (5), Shenzhen University (5), Shanghai University of Science and Technology (5), Tongji University (4) and other universities also have good performance.

Statistics according to the units to which the paper belongs.

Among the 1075 papers, nearly half of the papers were written independently by one unit, while more than half of the papers were written by three or four institutions.

Among them, 466 papers were completed independently by one unit, 377 by two institutions, 179 by three institutions, 37 by four institutions, and 13 and 3 by five to six units, respectively. Judging from the above data, the combination of academia and industry will become a trend.

Https://www.toutiao.com/i6752806374008160776/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.