Key Projects

Computational Theories and Methods for Brain-Machine-Integrated Perception and Cognition

Prof. Wu Zhaohui

Recent advances in the multidisciplinary fields of brain-machine interfaces, artificial intelligence, brain neuroscience, microelectronics, and neurophysiology signal a growing convergence between machine and biological intelligence. Both kinds of intelligence have their own intrinsic merits and demerits, but connecting and integrating them is an emerging trend. Cyborg intelligence aims to deeply integrate machine intelligence with biological intelligence by connecting machines and living beings via brain-machine interfaces, enhancing strengths and compensating for weaknesses by combining the biological cognition capability with the machine computational capability.

 

Cross-media Computing

Prof. Zhuang Yueting

Today there are lots of heterogeneous and homogeneous media data from multiple sources, such as news media websites, microblog, mobile phone, social networking websites, and photo/video sharing websites. Integrated together these media data represent different aspects of the real-world and help document the evolution of the world. Consequently, it is impossible to correctly conceive and to appropriately understand the world without exploiting the data available on these different sources of rich multimedia content simultaneously and synergistically.

Cross-media analysis is a research area in the general field of multimedia content analysis which focuses on the exploitation of the data with different modalities from multiple sources simultaneously and synergistically to discover knowledge and understand the world. Overall, cross-media analysis is beneficial for many applications in data mining, causal inference, machine learning, multimedia, and public security.

 

Analysis and Mining of Information Propagation on Social Networks

Prof. Cai Deng

In recent years, online social networks have undergone a rapid development, especially the birth and spread of blogs, forums and the twitter. This makes information propagation in social networks become a hot research area. Dr. Deng Cai, a professor of Computer Science department in Zhejiang University, is leading a group devoting to this area. Their research is focused on the structure of different social networks and how the information propagating on these networks. Their project is supported by the National Basic Research Program of China.

 

Multimedia Computing

Prof. He Xiaofei

Due to the rapid growth of the number of digital images, audios, and videos, there is an increasing demand for effective and efficient approaches for multimedia data representation, understanding, recognition and retrieval. Traditional content-based multimedia retrieval uses visual features instead of text features which facilitate the retrieval process. However, the semantic gap between low level visual features and high level concepts significantly reduces the retrieval performance. This project aims to develop novel approaches for semantic multimedia retrieval by using techniques from manifold learning and statistics.

 

Intelligent Media Computing

Prof. Cai Deng

Dr. Deng Cai has made major breakthrough in the theory of intelligent multimedia computing, multimedia recognition and understanding, and multimedia retrieval. He has published over 100 scientific articles in prestigious international journals and conferences in this area, and is widely recognized in the international community. His main contributions include: (1) He has studied the low dimensional representation problem of multimedia data by using machine learning theory, proposed several novel statistical analysis methods on data manifolds as well as a large scale manifold learning algorithm named spectral regression. (2) He has proposed several popular manifold-based face recognition algorithms, including Orthogonal Laplacianfaces, Neighborhood Preserving Embedding, Locality Sensitive Discriminant Analysis, and Semi supervised Discriminant Analysis. Combining with the tensor analysis of face data, he has built a complete manifold-based face recognition framework. (3)He has proposed a novel web page segmentation algorithm - VIPS, based on the visual cues. VIPS can divide a webpage into several semantic blocks. Based on VIPS, he has developed block-based web search algorithm, block-level link analysis algorithm and a novel method for searching web images and videos.

 

Data Visualization

Prof. Chen Wei

This project seeks to study visual analysis approach for unstructured and dynamic data. Analyzing and visualizing them has been a hot topic, which poses three key challenges: unstructured and heterogeneous data; dynamic data; and in-time analysis. The data to be handled denotes the ones that are produced with the rapid developmentof internet, mobile devices, sensor network and social network technologies, and demands rapid and insightful analysis, reasoning anddecision-making. Traditional data visualization targets visual display ofdata and user understanding of data patterns driven by visualization. Incontrast, this project aims to improve the performance and practicality of interactive data analysis assisted by data visualization. As such, the first goal is a novel data model that supports efficient visualization, analysis and interaction for unstructured and dynamicdata; the second part of this project is the technique and methods for enhancing existing data visualization technology; the third partfocuses on the data analysis framework within a visual analysis, reasoning and decision-making scheme. The efficiency of the proposed approaches will be evaluated by means of the OSINT data analysis.


Realistic Modeling and User Experience for Fashion E-commerce

Prof. Feng Jieqing

This project is funded by NSFC (Nature Science Foundation of China) key project program, led by Professor Feng Jieqing from our College. In this project, we will investigate the theories, methods, and key techniques of realistic modeling and user experiences in the whole process of fashion e-commerce. Meanwhile, we will investigate and implement the key techniques of e-commerce oriented personalized fashion design, freehand interaction-based virtual try-on, machine learning based fashion recommendation, size variable fitting robot, etc. Finally, we will develop a realistic try-on prototype system integrating both hardware and software. The research will be helpful to explore a new process of fashion e-commerce and provide the rigorous theoretical basis and technical support for fashion e-commerce upgrade in the Internet plus era.

 

Automatic Multi-modal Virtual Environment Construction Driven by Big Data

Prof. Xu Weiwei

This project is funded by NSFC key project program, led by Professor Xu Weiwei. The approved project organizes multi-channel perception information in an object-oriented manner, combines the virtual/augmented reality, big data and machine learning techniques to create novel techniques for virtual environment construction, including the construction of visual and auditory channel data set and body motion sensing dataset, hierarchical organization of environment knowledge, and the 3D reconstruction and automatic synthesis of virtual environments.

 

Efficient Scene Reconstruction and Rendering Engine Driven by Multi-source Data

Prof. Jin Xiaogang

The project is supported by the National Key R&D Program of China in 2017, led by Professor Jin Xiaogang. This project will investigate the challenges in virtual reality applications, such as scene reconstructions with low efficiency, low intelligence, and rendering very large-scale scenes in real time. The research will develop intelligent modeling software platforms and rendering engines for complex scenes by jointly considering the resource allocation between cloud and client, form a new methodology for them driven by multi-source data, and apply the developed technologies in demonstration applications such as e-commerce.

 

The Theories and Key Technologies of Crossover Services

Prof. Yin Jianwei

This project directed by Prof. Yin Jianwei was supported by the National Key R&D Program of China in 2017. The research on three key scientific problems will focus on service modeling, management and engineering methods. It was supposed to provide the solutions for the key technologies such as service pattern computing, crossover service design, integration and quality management. A supporting software platform will be developed and used in different areas in this project.


Intelligent Computer-aided Diagnose System Based on Fundus Images

Prof. Wu Jian

The project led by Prof. Wu Jian will focus on the follows: (1) simulating the way ophthalmologists interpret fundus images and utilizing Deep learning approach to determine the nature of lesions for automatic diagnosis of DR; (2) training the model with 88000 fundus images from EyePacs dataset and validating the model with labeled DR dataset;(3) training a Convolutional Neural Network with fundus images labeled by hospital for exudations and hemorrhages detection. This project will provide a methodology support for monitoring the disease and provides a scientific basis for DR early diagnosis and treatment.


Traffic and Road Sign Recognition and Understanding for Driverless Vehicles

Prof. Cai Deng

Being a novel technological innovation, the driverless vehicle system has a wide range of economic benefits and deep social impact. The proposed project aims to solve the difficulties in the driverless system, i.e, robust and accurate detection of road, vehicles, and traffic sign, pedestrian from the scene in realtime. We plan to develop the video analysis technologies with multi-view and multi-sensor channels. Also, we will explore 3D reconstruction, video segmentation, and multi-target tracking and recognition in complex real-time weather conditions, which are the fundamental problem in computer vision. These technologies will provide sustainable development of the theory and technical support forrobots, intelligent transportation, video surveillance, and etc.

 

Study on the Fundamental Theory and Methods of Exploratory Visual Analysis

Prof. Chen Wei

This project will study a new exploratory data analysis mode by integrating visualization, data mining in a visual interface: visual analysis. By leveraging the human vision perception capability, visual analysis incorporates human intelligence, especially the perceptible but inexpressible knowledge and personalized experience into the entire data analysis and decision making process within the visual interface. Thereby the data complexity is progressively decreased to the degree that can be processed by human and computer, yielding useful knowledge. This project will take practical applications as examples to study and evaluate new schemes of exploratory visual analysis, whose goal is to iteratively achieve the mutual promotion of human intelligence and machine intelligence through visual interface to obtain new knowledge. Each stage in this steering procedure has associated new challenge. The project will emphasize on three aspects: 1) how to define and construct easy- to-visualize, mining and interaction model for specific application goals; 2) how to efficiently accomplish analysis-driven visualization; and 3) how to fulfill the interaction and fusion of human users and machine intelligence in an integrated visual analysis environment.

 

Geometric Optimization and Dynamic Simulation of Complex Physical Systems

Prof. Bao Hujun

Simulation technology has been widely used in engineering, production, scientific research, entertainment, virtual reality and other fields. However, as involved in a variety of applications, the physical object and the increasing complexity of the simulation process involves extremely complex algorithm design and huge computational cost, current computer simulation technology faces enormous challenges in the accuracy of the simulation results, the computational efficiency and control which has become the bottleneck of many applications. Complex physical objects manifest itself complex ingeometry and governing equation. According to this, most simulation of complex physical objects can be regarded as solving complex partial deferential equation on complex domain. For efficient calculation, appropriate transformation, approximation and discretization of equations become crucial. This project takes complex physical object as research target, takes efficient and controllable simulation process as a goal, and conducts research by optimizing the geometric modeling method, representation and discretization strategy. This study is significant to theory and methods in physical simulation. Furthermore, it will provide strong technical support for a variety of applications.

 

China Knowledge Centre for Engineering Science and Technology

Prof. Pan Yunhe, Prof. Zhuang Yueting

China Knowledge Centre for Engineering Sciences and Technology (CKCEST) was launched by Chinese Academy of Engineering (CAE) in March 2012 and plans to bring together the huge amounts of data in the field of engineering science and technology in about 9 years to construct research service systems in the field of engineering science and technology. The goal is to construct an integrated body of knowledge with the most abundant information, the most extensive range of applications, and the most pragmatic nature. The research team at Zhejiang University School of Computer Science and Technology has participated in the CKCEST project as a major force. To gather dispersive data so as to build databases with easy accessibility and generation of new knowledge is the core technology of CKCEST project. CKCEST is to make massive knowledge more easily accessible with technological innovation.

 

Self-Driving Smart Car Platform and Key Technologies

Prof. Wu Zhaohui

A Smart Car is an intelligent wheeled mobile robot, combining Wireless Sensor Networks (WSN), big data, mobile internet and intelligent manufacturing. Self-driving cars are the future direction of Smart Cars. This project focuses on key technologies including environment perception based on computer vision, obstacle detection based on radars, the Internet of Vehicles, electric vehicles based on new energy sources, etc. The project aims to develop intelligent vehicle products, to help address the difficult issues such as traffic jams, air pollution and energy shortage in Chinese cities, and promote the development of intelligent transportation industry in China. The self-driving electric mini-cars developed by the project integrate multiple functionalities, including environment perception, speech control, and path planning and obstacle avoidance. The first generation of intelligent min-cars has been put into practical use for people to rent on university campuses and in city parks.

 

Theories and Methods of Digital Protection and Utilization of Cultural Heritage

Prof. Lu Dongming

The digital protection and utilization of cultural heritage is one of the core content of national culture power strategy, its research will promote further integration of culture and science and technology, and improve the national soft power. With the application of digital technology in the field of cultural relics protection, the digital protection and utilization of cultural heritage gradually formed a new cross subject, and its connotation and denotation is expanded and enriched. According to the demand of the cultural heritage protection and utilization, this direction consists of confition monitoring and environmental awareness, digital collection and record, multivariate information extraction, digital resourcesprocessing, data mining and analysis, simulation. The system had been used in 11 provinces throughout the county.

 

Rehabilitation Service for People with Disabilities

Prof. Bu Jiajun

This project constructs a nationwide and accessible rehabilitation service platform. Building upon a series of rehabilitation oriented technologies including rehabilitation service sensing and recommendation, rehabilitation data collecting and exchange, rehabilitation information retrieval, remote rehabilitation guidance, rehabilitation data analysis and mining etc., theplatform provides realtime online rehabilitation service to the disabledpersons nationwide. By integrating data and services from the third-party assistive device systems, the platform supports effective resource sharing and serves various stakeholders including disabled persons, rehabilitation experts, assistive device manufacturers; professional rehabilitation institutes etc. The platform is now online via the ChinaDisabled Persons Service Network for nationwide demonstration and has achieved significant socioeconomic performance.

 

Health Services Platform and Application Development for the Aged

Prof. Yin Jianwei

This project proposes a collaborative pattern to achieve a promising solution for the health service of the aged in China. We developed a health service platform composed of three major parts: interface and detecting hardware, healthcare service and back-end software. Health detecting hardware includes home tablets, tablets for community doctors, four-in-one body health surveillance, and bioelectricity-based sub-health surveillance devices. Service is the key of the whole system, offering aged oriented professional service including health consultancy, chatting, observation, touching, searching, entertainment, emergency call and normal service. Also, the project group carries out demonstration and application in more than five million families, one hundred communities, and ten hospitals in Hangzhou, Fujian, Dalian, Hunan and Beijing.

 

PaaS Cloud Platform

Prof. Yang Xiaohu

Instead of the virtual compute and storage provided by most cloud service providers, Platform-as-a-Service (PaaS) cloud technology provides a higher level of service by managing applications, middleware and software runtime environment. Currently IBM, Redhat, SAP, HP are investing heavily in PaaS. VLIS lab of Zhejiang University started researching PaaS related technologies in 2011, and we have developed a self-owned distribution of PaaS cloud platform: zFoundry. We share our R & D work at top PaaS clod computing summits and have successfully used zFoundry to support ‘smart city management’ in Hangzhou and ‘smart food and drug monitoring’ in Tianjin.

 

Large-Scale Mission Critical System Development

Prof. Yang Xiaohu

Large-Scale Mission Critical System refers to the system in which downtime caused by any factors will result in operation failure. This system with extremely complicated mission has a high demand on precision, performance, scalability, availability, reliability and maintainability. Since 2001, VLIS lab of Zhejiang University has focused on pioneering development of Large-Scale Mission Critical System with profound accumulation of distribution technique, high reliability technique and high performance technique. In long-term cooperation with many internationally famous companies such as State Street, Cisco and so on, we have developed many core financial systems including Large-Scale Global Foreign Exchange System, System of Compliance Inspection and Valuation on Mutual Fund.

 

Style Image-oriented Modeling and Design of Product Family Shape Gene

Prof. Luo Shijian

This project will explore style image-oriented modeling and design of product family shape gene to shape the unique product image, in which the theory of gene in biology with inheritance and variation was borrowed, and the knowledge with industrial design, computer technology and psychology was integrated. Taken the car shape design as an example, this project will (1) define the relationship between product family shape feature and style image, and build the three-level representation system of style image, feature and shape gene; (2) set up explicit relationship between currency gene, adaptable gene and individual gene and establish the semantic expression model of product family shapegene; (3) create the mapping relation between style image and the semantic model of product family shape gene; (4) employ neural networks and multi-user interactive genetic algorithm to build intelligent learning and reasoning mechanism to establish style image-oriented product family shape gene generating system, which can realize the bi-directional reasoning between style image and design. This project will present new theories and methods tor product innovation design from design methodology, which can not only provide guidance for enterprises product development and branding, but also promote the development of industrial design and knowledge engineering.

 

Creative Design-oriented Culture Heritage Knowledge System Study and Knowledge Base Construction

Prof. Luo Shijian

This project is led by Zhejiang University and jointly undertaken by Tianjin University, Beijing ITEI Measurement & Control Technology Co. Ltd and Beijing Ying Ke Da Cheng Science and Technology Ltd. The whole project is all with the goal of Research on key technology and application of the analysis of cultural relics and recreating design material. Firstly, this project will investigate the history, art and culture from form, constructure and pattern within four culture heritages (the utensils, fabrics, architectureand frescoes), and explore related knowledge system and acquisition methods forcreative design; Secondly, this project will set up culture heritage knowledge databaseand provide a sound basis for cultural creative design; Finally, by investigating the technology of cultural component testing and data analysis, exploration on culturalknowledge and cultural elements, construction of cultural knowledge base and creative material library, this project develops a platform for modern creative design and construe the value chain of cultural creative industry.

 

Strategy for Innovation Design

Prof. Lu Yongxiang, Prof. Pan Yunhe, Prof. Sun Shouqian

Development Strategy for Innovation Design is a major consulting project supported by Chinese Academy of Engineering. The goals of the project are to fully understand the significance and function of innovation design in manufacturing industry as well as in the development of social economy in our country. To speed up the implementation of the national transformation from the big manufacturer to the powerful manufacturer, and to further boost the transformation and upgrading of building an innovation-oriented country. Based on innovation design and manufacturing face to the overall economic and social situation, the research will not only provide a policy suggestions and decision-making basis for government to formulate development strategy and planning, but also enhance the innovative design ability by research, education manufacture. Media, consumer and finance, promote development of innovation-driven design and manufacturing to make all people value, respect and support innovation design. Supported by the project, on Oct 11, 2014, the Innovation Industry Strategic Alliance of China (IDAC) was founded and located (its secretariat) in Zhejiang University, and the news was broadcasted by CCTV 1.


Realistic Modeling and User Experience for Fashion E-commerce

Prof.Feng Jieqing

This project is funded by NSFC (Nature Science Foundation of China) key project program, hosted by Professor Feng Jieqing from our College. In this project, it will investigate the theories, methods, and key techniques of realistic modeling and user experiences in the whole process of fashion e-commerce. Meanwhile, it will investigate and implement the key techniques of e-commerce oriented personalized fashion design, freehand interaction-based virtual try-on, machine learning based fashion recommendation, size variable fitting robot, etc. Finally, it will develop a realistic try-on prototype system integrating both hardware and software. The research will be helpful to explore a new process of fashion e-commerce and provide the rigorous theoretical basis and technical support for fashion e-commerce upgrade in the Internet plus era.



Automatic Multi-modal Virtual Environment Construction Driven by Big Data

Prof.Xu Weiwei

This project is funded by NSFC key project program, hosted by Professor Xu Weiwei. The approved project organizes multi-channel perception information in an object-oriented manner, combines the virtual/augmented reality, big data and machine learning techniques to create novel techniques for virtual environment construction, including the construction of visual and auditory channel data set and body motion sensing dataset, hierarchical organization of environment knowledge, and the 3D reconstruction and automatic synthesis of virtual environments.



Efficient Scene Reconstruction and Rendering Engine Driven by Multi-source Data

Prof.Jin Xiaogang

The project is supported by the National Key R&D Program of China in 2017, hosted by Professor Jin Xiaogang. This project will investigate the challenges in virtual reality applications, such as scene reconstructions with low efficiency, low intelligence, and rendering very large-scale scenes in real time. The research will develop intelligent modeling software platforms and rendering engines for complex scenes by jointly considering the resource allocation between cloud and client, form a new methodology for them driven by multi-source data, and apply the developed technologies in demonstration applications such as e-commerce.


The Theories and Key Technologies of Crossover Services

Prof.Yin Jianwei

This project hosted by Prof. Yin Jianwei was supported by the National Key R&D Program of China in 2017. The research on three key scientific problems will focus on service modeling, management and engineering methods. It was supposed to provide the solutions for the key technologies such as service pattern computing, crossover service design, integration and quality management. A supporting software platform will be developed and used in different areas in this project.


Intelligent Computer-aided Diagnose System Based on Fundus Images

Prof.Wu Jian

The project hosted by Prof. Wu Jian will focus on the follows: (1) simulating the way ophthalmologists interpret fundus images and utilizing Deep learning approach to determine the nature of lesions for automatic diagnosis of DR; (2) training the model with 88000 fundus images from EyePacs dataset and validating the model with labeled DR dataset;(3) training a Convolutional Neural Network with fundus images labeled by hospital for exudations and hemorrhages detection. This project will provide a methodology support for monitoring the disease and provides a scientific basis for DR early diagnosis and treatment.



Copyright © 2018 College of Computer Science and Technology, Zhejiang University