Recently, the Applied Mathematics and Artificial Intelligence Research Team of the Department of Mathematics, College of Sciences, presented a paper titled SingleImage Super-Resolution via a Holistic Attention Network at the European Conference on Computer Vision 2020 (ECCV2020), an international top-level conference on computer vision. The paper is jointly completed by the Applied Mathematics and Artificial Intelligence Team of NEU, Xidian University, Chinese Academy of Sciences, Australian National University, and Didi Artificial Intelligence Research Institute. NEU is the first completion unit of the paper, Niu Ben, a graduate student of 2018 from the Applied Mathematics and Artificial Intelligence Team of the Department of Mathematics, College of Sciences, is the first author, and associate professor Yang Lianping from the Department of Mathematics, College of Sciences, is the corresponding author.
This paper presents a new holistic attention network (HAN). This network consists of a layer attention module (LAM) and a channel spatial attention module (CSAM). LAM is used to consider the correlation between layers so as to adaptively emphasize the characteristics of layers. At the same time, CSAM is used to learn per-pixel confidence at all locations of each channel to selectively capture more information characteristics. The modeling of the research results is carried out based on the overall interdependencies among layers, channels and positions, thus obtaining a single image super-resolution method superior to others.
Niu Ben studied at NEU for an undergraduate degree. In 2018, he was recommended to pursue a master's degree in applied mathematics at NEU without taking an examination. In September 2018, he joined the Applied Mathematics and Artificial Intelligence Team of the College of Sciences and began to carry out in-depth research on machine learning and computer vision. Adhering to the talent cultivation philosophy of turning mathematics into technology and data into knowledge, the Applied Mathematics and Artificial Intelligence Team of the College of Sciences has been paying attention to integrating theoretical learning, knowledge application and skill improvement to comprehensively improve its talent cultivation quality. Its graduates are widely employed by Huawei, Tencent, Alibaba, Baidu, Jingdong and other famous enterprises, and are favored by the employers. Huawei has provided scholarships for the graduates of the team for three consecutive years, and Niu Ben is one of the recipients of Huawei Scholarships in 2019.
Computer vision is a comprehensive discipline covering applied mathematics and statistics, computer science and engineering, signal processing, physics, neurophysiology and cognitive science. ECCV, held every two years, is recognized as one of the top three conferences in the field of computer vision. The main collected papers are all from the world’s top laboratories and research institutes focusing on theoretical research.