Multimedia Institute


The Multimedia Institute (M2I) in the department of Electronic Information Engineering, Tianjin University, is formerly known as the Institute of Television and Image Information approved by the Ministry of Education of the People’s Republic of China. M2I is dedicated to research of multimedia content understanding, retrieval, security and communication.

Our research interests widely cover the related problems in computer vision, machine learning, signal processing and information retrieval and particularly focus in the following areas:
1. Computer Vision And Machine Learning
2. Biomedical Image Analysis
3. Multimedia Understanding And Retrieval
4. Media Security And Forensics



  1. 有浓厚的科学研究兴趣,希望未来从事相关领域技术研究;
  2. 有浓厚的系统开发兴趣,希望未来从事相关行业技术开发;
  3. 有创新创业精神,有志于尝试在领域内进行创业创新;
  4. 985、211高校或有研究生院的高校的推荐免试研究生优先考虑。


  1. 推荐优秀博士生和硕士生赴国内及国外合作团队进行联合培养;
  2. 提供优秀博士生和硕士生参加顶级国际会议的资助;
  3. 提供相应的科研开发环境和工程应用项目支撑;
  4. 对于有创业能力和背景的研究生提供技术、团队、资金、场地等支持;
  5. 提供课题组间的跨学科、跨领域交流,并邀请国际著名学者来实验室交流。

Dr. LIU An-An will serve as the co-chair of the workshop on Computer Vision for Microscopy Image Analysis, which is in conjunction with CVPR 2016 conference.

Our lab organized the view-based 3D model retrieval competition in the SHREC workshop of Eurographics 2016.  [Project]

One paper, "3D Convolutional Networks-Based Mitotic Event Detection in Time-Lapse Phase Contrast Microscopy Image Sequences of Stem Cell Populations", was accepted by CVPR 2016 workshop. This paper proposed a straightforward and effective method for mitotic event detection in time-lapse pahase contrast microscopy image sequences of stem cell populations.

Deep learning on TJU painting.  [Project]

Two master students begin the internship in SeSaMe of NUS from 09/2015 to 09/2016.

One paper, "Clique-graph Matching by Preserving Global & Local Structure", was accepted by CVPR2015. Liu An-An and Nie Wei-Zhi attended the conference. [Project]

Our lab organized the view-based 3D model retrieval competition in the SHREC workshop of Eurographics 2015. 6 participants submitted 13 runs. The comparions demonstrated that MV-RED is a challenging benchmarf for the evaluation of 3D model retrieval.  [Project]