China's highest award for intelligent science and technology - the "2022 Wu Wenjun Artificial Intelligence Science and Technology Award" award ceremony was held in Beijing today. More than a hundred experts from ministries and commissions, domestic and foreign academicians, top scholars, award-winning representatives and entrepreneurs attended. Researcher Xing Junliang from Tsinghua University, Li Jianshu and Zhao Wenbiao from Ant Group won the first prize of the "Natural Science Award" for the achievement of "Unconstrained Portrait Target Intelligent Perception and Understanding".
The Wu Wenjun Artificial Intelligence Science and Technology Award was established by the Chinese Association for Artificial Intelligence in 2011. It is the first award named after "artificial intelligence" in Chinese history. The first prize and above are qualified to nominate and recommend the National Science and Technology Award. In the field of artificial intelligence in my country Enjoy the highest honor symbol. Among them, the "First Prize of Natural Science Award" is limited to those who have achieved major breakthroughs in forward-looking basic research, leading original achievements, recognized and widely cited by the scientific community at home and abroad, promoted the development of disciplines, or had significant impact on the economy and society. .
This time, the "Intelligent Perception and Understanding of Unconstrained Portrait Objects" that Tsinghua University and Ant Group completed over ten years won the "First Prize of Natural Science Award", which means that my country has made great achievements in solving the core problems in the field of AI vision. Innovative breakthrough.
(Picture: On May 6, the 2022 Wu Wenjun Artificial Intelligence Science and Technology Award Ceremony, Xing Junliang of Tsinghua University, Li Jianshu of Ant Group and others jointly completed the "Unconstrained Portrait Target Intelligent Perception and Understanding" achievement won the "Natural Science Award 1st Prize")
Artificial intelligence has become an important engine for a new round of technological revolution and industrial transformation. As an important part of the artificial intelligence industry, computer vision gives machines the ability to "see" and "cognition" by simulating the human visual system, which is a key technology for realizing industrial digitalization and intelligence. Portrait objects are the most important research content in the field of computer vision, and have a wide range of application requirements in financial authentication, people's livelihood and economic scenarios. However, due to the internal variable factors of portraits (such as posture, expression, gender, etc.) and external complex environmental factors (such as viewing angle, illumination, noise, etc.), the existing portrait target perception and understanding models have great challenges in accuracy and efficiency. .
In response to this complex scientific problem, after more than ten years of hard work, the project has formed a relatively complete basic theoretical method and core algorithm technology research system for the intelligent perception and understanding of portrait targets. There are three main scientific discoveries: First, it reveals the importance of "invariant feature learning" for the changeable appearance and shape of portraits, and forms a theoretical framework for perceptual modeling of key information of full-view portraits. The second is to focus on the intrinsic correlation attributes of portraits in complex spaces, and find that there are common correlations between multi-dimensional attributes such as posture and expression of portraits, and propose an attribute mining method of "joint multi-task and confrontational invariance learning". The third is to combine the key information and related attributes of portraits to construct a highly applicable machine vision intelligent technology application framework for "face-body-crowd", which can solve problems such as multi-attribute fusion perception in actual application scenarios.
It is understood that with the support of the National Natural Science Foundation of China and the "New Generation Artificial Intelligence" major project, the project has achieved fruitful results in academic research, technological impact, and service industries. Has won 5 important paper awards such as CVPR and ACM Multimedia, and won 10 championships in domestic and foreign top portrait perception and understanding competitions; 8 representative papers have been cited 3225 times by Google-Scholar, 650 times by WOS, and won more than ten places The winner of the Marr Prize/Fu Jingsun Prize, more than 100 domestic and foreign academicians and other well-known scholars quoted and positively commented.
The relevant results of this project have been applied in the core products of enterprises such as Omron, Huawei, and Ant Group, and have produced large-scale commercial applications. Ant Group's trusted digital identity authentication solution based on the project's core capabilities has served more than 120 million Southeast Asian users overseas, and won the IDC Financial Technology Real Value Award.
