Intelligent Sensing Foundation
	The goal of intelligentsensing group is to understand the underlying learning principles delivering intelligence and emulate the perceptual capability of human. It spans the spectrum of topics in image and signal processing, computer vision, pattern recognition and learning, reconstruction and rendering, from mathematical theory to practical applications and from low-level sensing to high-level understanding. It undertakes projects from National Basic Research Program of China, National Natural Science Foundation of China and other research projects. Major research interests of the intelligentsensing group lie in two-folds: 1) Learning theory and optimization methods that endow machines with ability to surpass and assist the human vision system in certain ways. The current research topics include ordinal measure theory, half-quadratic optimization, sparse representation learning, information theoretic learning, multi-modal learning, and deep learning. 2) The acquisition, processing, analysis and rendering of visual information. It studies computational photography methods to efficiently capture and process visual data. It combines cognition mechanisms and machine learning methods to analyze and simulate light for emerging computer vision areas, e.g., generative models, transfer learning, self-supervision.
Biometrics and Forensics
	The research of biometrics and forensics is driven by application requirements in public security, Internet, Internet of Things, artificial intelligence industry, etc. We focus on the key problems in acquisition and recognition of iris and face biometrics, and visual and acoustic content authentication. The research objective of biometrics is to improve robustness of iris and face recognition in real-world applications using novel imaging and algorithms. And the objective of forensics is to analyze the integrality and authenticity for the image and video content based on pattern recognition and statistical analysis tools. Our research has been supported by the National Basic Research Program of China, the National Hi-Tech Research and Development Program of China, the National Key Technology R&D Program, the National Natural Science Foundation of China. Our research on iris recognition was awarded the National Award for Technological Invention 2nd Prize. Furthermore, we have released large scale databases on multi-modal biometrics and image forensics which have been used by more than 30,000 research groups from more than 170 countries. There are totally three spin-off companies in this direction, with successful applications in public security, banking, etc.
Biologically-Inspired Intelligent Computing
	The BRAVE group is on a scientific mission to push the boundaries of AI, developing novel AI theories and methods to achieve the breakthrough of current technical limitations, i.e. robustness, adaptability, interpretability etc. The group tries to seek inspirations from the biological mind, i.e. neural architectures, cognitive mechanisms, learning modes as well as evolution trajectories, to solve the most pressing real-world challenges on multi-source data fusion, multi-sensor vision measurement, open-set semantic perception and understanding and Human-machine Symbiosis and Hybrid Intelligence.The group has three academic areas: 1) Bio-inspired machine learning, trying to construct human-like learning theories enable to solve problems without needing to be taught how. 2) Open-set semantic perception and understanding, trying to simulate the cognitive nature of human brain and break the bottlenecks of open-set semantic perception and understanding, i.e. lacking of labeled data, weakly supervised information, unknown categories and variable distributions. 3) Autonomous evolution of machine intelligence, trying to imitate evolution trajectories of the biological mind, arouse the artificial agents to enhance their intelligence autonomously. Until now, the group has published over 100 peer-reviewed papers on the prestigious journals and the mainstream conferences, and won the several influential awards, include best paper award of BICS2016. It undertakes projects from National Key R&D Program of China, National Natural Science Foundation of China, Military Equipment Pre-research Program and other research projects.In the future, the group will continue to pursue the positive and transformative innovations of AI and provide advanced AI techniques for daily life, intelligent manufacturing, etc.
Multimodal Intelligent Computing
	The multi-modal intelligentcomputing research group deals with different types of data, such as images, texts, videos, etc. It studies both theory and applications about pattern recognition, visual computing, machine learning, data mining, context modeling, etc. Major research interests of the multi-modality computing group include: (1)Multi-modal intelligent analysis technology based on deep learning. Research of methods and applications on fusion methods of multi-modal data, e.g., image, text and speech, based on deep learning, cross-modal retrieval, cross-modal data generation.(2) Methods and applications for visual computing based on deep learning. It studies how to efficiently integrate the feedback mechanism in feedforward networks and how to combine active vision in feedforward and feedback networks, which can be used to solve many vision tasks in large scale vision analysis, e.g., object recognition, object detection, video segmentation and video analysis. (3)Visual surveillance analysis for public safety. The intelligent analysis requirements of massive video for big data environment, research on key technologies such as target detection, motion tracking, attribute recognition, cross-scene re-identification, action-behavior-event recognition, and establish a data analysis platform solves the difficult problems such as massive target retrieval and abnormal behavior detection. (4)Intelligent data analysis for public security and business intelligence. It studies the key technology of large scale social network data mining taking advantage of advanced technologies of big data, such as context modeling, time-series prediction and user modeling, which adapts to the needs of public and contentsecurities.