ZEEWAIN - Tsinghua AI-enabled Platform Created Good Results in the "Science Innovation Competition"

The industrial implementation of the pedestrian re-identification (ReID) technology of the team of Professor Wang Shengjin of the Department of Electronic Engineering has achieved important progress. ZEEWAIN Technology, the enterprise with the technology achievement commercialization as the mainstay, after three months of three rounds of selections level by level, stood out from 2625 enterprises that participated in the competition, and won awards in the 9th China Innovation & Entrepreneurship Competition (Guangdong • Guangzhou Division) and the 5th Guangzhou "Science Innovation Cup" Innovation & Entrepreneurship Competition Finals (New Generation of Information Technology Industry Growth Group). The competition was jointly organized by Ministry of Science and Technology, Ministry of Finance, Ministry of Education, Cyberspace Administration of China, and All-China Federation of Industry and Commerce. The number of enterprises that registered for the preliminary competition in Guangzhou Division increased by 35% compared with that in the last year. After the quarter-final of 865 enterprises, the semi-final of 562 enterprises, and the final of the top 8 enterprises in each industry, finally, a total of 48 top scientific enterprises from 6 industries stood out and entered the grand final.

Pedestrian ReID is currently recognized as a challenging frontier research topic in the computer vision research field. The target produces time varying and deformation at the same time with important theoretical research and application value. There are problems in this research topic such as non-rigid and non-cooperative target, large-scale internal variance, small training sample, high generalization ability rudiments, large deformation, large changes in viewing angle, and serious lighting effects. Targeted at the above-mentioned main problems, Professor Wang Shengjin has led his team to conduct in-depth research on the pedestrian identification characteristics expression theory and methods for nearly 10 years. The research in this field is at the forefront of the world and some results take the leading position. Meanwhile, the breakthrough in pedestrian ReID technology preliminarily solves the bottleneck problem of pedestrian target tracking with cross-view cameras, and strongly promotes the application of intelligent video security.

Many representative papers of pedestrian ReID results of the team of Professor Wang Shengjin have been published in important international journals and top conferences in this field such as PAMI, CVPR, ICCV, and ECCV, and the academic citations of pedestrian ReID related representative works in Google Scholar have exceeded 3,000 times with the citations of single ones of over 1,500 times. The results have been cited by relevant course materials of well-known foreign universities and have a relatively big influence in international counterparts. Among them, in the research of pedestrian ReID visual characteristic expression, the SVDNet method was proposed, using SVD (singular value decomposition) interpretation and optimization of in-depth network model to achieve the orthogonal constraint of discrimination. The paper was selected and published as a spotlight by the top conference ICCV2017 in this field; one paper was evaluated by the International Association for Pattern Recognition as the ICPR2018 Best Scientific Paper, and one paper was evaluated as "The Best of the Physics arXiv" from February 14 to February 21, 2015 by MIT Technology Review, and one doctoral dissertation was evaluated as the 2017 Excellent Doctoral Dissertation of the Chinese Association for Artificial Intelligence. The team of Professor Wang Shengjin has constructed and published three large-scale public data sets in the pedestrian ReID research field, which have become the standard test data sets for the research in this fields, with a total downloads of over 13,600 times. The research results won the 2019 Wu Wenjun Artificial Intelligence Science and Technology Award and the Science and Technology Award of the Ministry of Public Security respectively. ZEEWAIN Technology, a Tsinghua technology industrialization company with this technology as the mainstay, has the world's leading AI algorithm intelligence -- pedestrian ReID • multi-modal algorithm, vehicle ReID algorithm, brain-like intelligence -- small sample machine learning, image understanding, cross-media intelligence and other series of core technologies, some of which have been applied and implemented in corresponding industries.

The construction of the Guangdong-Hong Kong-Macao Greater Bay Area is a national strategic plan and is regional planning of the century focused by the whole world. "Let more enterprises have AI innovation capabilities! Continuously lead the development of the AI technology and enable people to continuously enjoy a better life of intelligent innovation" is the vision of the Tsinghua technology industrialization company ZEEWAIN Technology. It has massive flows of people, vehicles, logistics and data, with a foundation of rich AI application scenarios, being the highland for China to promote the innovative application of new technologies such as AI and build the world's digital economy. Meanwhile, the Ministry of Science and Technology also wrote to the People's Government of Guangdong Province to support Guangzhou to construct a national new generation of AI innovative development pilot zone. This series of planning and basic resources have provided a huge market space and industrial ecological chain for ZEEWAIN, which can fully support ZEEWAIN to take root in the Guangdong-Hong Kong-Macao Greater Bay Area, research the AI technology well, do a good job of the innovative application of AI in all walks of life together with partners, and strive to become a world-class AI-enabled platform and enterprise.

Contact: Wang Shengjin: wgsgj@tsinghua.edu.cn