Dr. Juan Cao (²Ü¾ê)


enjoy my work, enjoy my life

MultiMedia Computing Group
Institute of Computing Technology
Chinese Academy of Science

TEL: +86-010-62600659   Email:caojuan@ict.ac.cn

No.6 South Kexueyuan Road, Haidian District, Beijing 100190, China

Biography
      I am an assistant professor at Multimedia Computing Group (MCG), Institute of Computing Technology (ICT), Chinese Academy of Science (CAS), the leader of MCG is Prof. YongDong Zhang. I received my Ph.D at ICT supervised by Prof. JinTao Li, and I got the B.E. and M.S. degrees in computing science from Xiang Tan University supervised by Prof. JingYe Zhou.
          I was a Senior Research Associate in the Video Rtreival Group(VIREO), City University of Hong Kong from May 2009 to August 2009, under the supervision of Dr. Chong-Wah Ngo.

Research Interests

  • Multimedia information retrieval
  • Large scale web video mining
  • Machine learning, Topic model

    Systems and Databases

    I have participated many large projects. As the person in charge, I have successfully designed and developed the following video retrieval systems with my excellent team members.

  • Interactive video retrieval system: VideoMap

    • Here we introduce the highlights of our interactive video retrieval system VideoMap. To enhance the efficiency, the system has a map-based displaying interface, which gives the user a global view about the similarity relationships among the whole video collection, and provides an active annotating manner to quickly localize the potential positive samples. Meanwhile, the proposed map supports multiple modality feedback, including the visual shots, high-level concepts and keywords. The system can improve the retrieval performance by automatically optimizing these feedback strategies.



                 Figure 1 VideoMap¡¯s retrieval interface

            Figure 2 video map for active recommendation

    The demos of VideoMap system can be found in http://mcg.ict.ac.cn/chengguo2.html


  • Large-Scale Web Video database: MCG-WEBV

    •         The goal of MCG-WEBV is constructing a general database for all kinds of web video research. Firstly, it collects the most viewed videos of every month on YouTube, which are most valuable to do mining for their high quality and popular contents. Meanwhile, the database is expanded to the related videos and ones uploaded by the same authors, which aims to keep the original social network information on YouTube. Secondly, the database provides comprehensive features for the video analysis and process. It includes the raw videos, keyframes, five metadata features, eight web features, and eleven low-level features cover textual, visual and audio. Finally, the database also includes the ground-truth of 73 hot web topics by human annotation, and the labels of 15 video categories from YouTube website. Many applications such as topic discovery and track, web video categorization etc. can be implemented and evaluated on this dataset, and we believe that MCG-WEBV will provide a significant foundation for web video research.
            The version1.0 of MCG-WEBV consists of 80031 videos from Dec. 2008 to Feb. 2009, with 3283 core videos and 76748 expanded videos. The database is updated every month.

    The database can be downloaded from http://mcg.ict.ac.cn/chengguo1.html

  • Automatic video retrieval system in TRECVID



    • Figure 3. the flowchart of MCG-ICT-CAS automatic search system
            The MCG-ICT-CAS automatic search system for TRECVID 2008 is as Fig. 3. In the concept-based module, we propose a novel distribution based concept selection (DBCS) approach, which achieves a stable good performance for all the topics (0.053). In the visual-based module, we focus on the low dimensional semantic features by Latent Dirichlet Allocation model and get an infAP of 0.033. Finally, a re-ranking technology based on the motion and face and a multi-runs and multi-examples fusion approach (SSC) were applied to aggregate the basic search results, which produced a significant improvement.

    We participate the search task of TRECVID from 2007 to 2009, and have got encouraging results with the first-rank in 2008 and the second-rank in 2007. The system performance of 2009 is under evaluation yet.


    Figure 4. MCG-ICT-CAS automatic search
    results in TRECVID 2008


    Figure 5. MCG-ICT-CAS automatic search
    results in TRECVID 2007

  • Go2View web video retrieval engine


    •         This is a prototype system for web video index and retrieval. It includes the collector for web video content, high-dimensional indexing technology, and multi-modality retrieval supporting the visual examples and textual queries.
            My main work in this system is the text-based index and retrieval. Moreover, I developed a semantic retrieval module, which can automatically analyze the queries by Latent Semantic Indexing(LSI),and find the semantic relevant results.

    Publications

    2009

    1. Juan Cao, HongFang Jing, Chong-Wah Ngo, YongDong Zhang, Distribution-based Concept Selection for Concept-based Video Retrieval, ACM International Conference on Multimedia (ACM MM), Beijing, China, Oct. 2009.
    2. Juan Cao, Tian Xia, Jintao Li, YongDong Zhang, Sheng Tang, A density-based method for adaptive LDA mod el selection, Neurocomputing, 72(7-9): 1775-1781 (2009)
    3. Juan Cao, YongDong Zhang, JunBo Guo, Lei Bao, JinTao Li, VideoMap: An Interactive Video Retrieval System of MCG-ICT-CAS. ACM International Conference on Image and Video Retrieval (CIVR), Santorin, 2009.
    4. Juan Cao, YongDong Zhang, YiCheng Song, ZhiNeng Chen, Xu Zhang, and JinTao Li, MCG-WEBV: A Benchmark Dataset for Web Video Analysis, Technical Report, ICT-MCG-09-001, Institute of Computing Technology, May. 2009
    5. Tian Xia, Juan Cao, YongDong Zhang, and JinTao Li. On defining affinity graph for spectral clustering through ranking on manifolds. In Neurocomputing , 72 (2009) 3203-3211.
    6. Lei Bao, Juan Cao, Tian Xia, YongDong Zhang, JinTao Li, Locally Non-negative Linear Structure Learning for Interactive Image Retrieval, ACM International Conference on Multimedia (ACM MM), Beijing, China, Oct. 2009.
    7. BaiLan Feng, Juan Cao, ShouXun Lin, YongDong Zhang, Kun Tao, motion region-based trajectory analysis and re-ranking for video retrieval, IEEE International Conference on Multimedia and Expo(ICME£©, 2009
    8. Xu Zhang, Yicheng Song, Juan Cao, Yongdong Zhang, Jintao Li, "Large Scale Incremental Web Video Categorization", In Proceeding of the 1st ACM MM2009 Workshop on Web-Scale Multimedia Corpus (WSMC), Beijing, 2009.
    9. YiCheng Song, YongDong Zhang, Xu Zhang, Juan Cao, JinTao Li. Google Challenge: Incremental-Learning for Web Video Categorization on Robust Semantic Feature Space. In Proceedings of the 17th International ACM Conference on Multimedia (MM2009), Beijing, China, November 2009.

    2008

    1. Juan Cao, Jintao Li, Yongdong Zhang, The optimal condition of LDA model for video retrieval, Chinese Journal of Computers.Vol.31, no.10, pp.1780-1787,2008.
    2. Juan Cao, Yongdong Zhang, Bailan Feng, Xiufeng Hua, Lei Bao, and Xu Zhang, MCG-ICT-CAS TRECVID2008 search task report, TREC Video Retrieval Evaluation Online Proceedings (TRECVID), 2008

    2007

    1. Juan Cao, Jintao Li, Yongdong Zhang, Sheng Tang, LDA-Based Retrieval Framework for Semantic News Video Retrieval.IEEE International Conference on Semantic Computing (ICSC), 155-160, 2007
    2. Juan Cao, Sheng Tang, Jintao Li, Yongdong Zhang, Xuefeng Pan, A Lexicon-Guided LSI Method for Semantic News Video Retrieval, PCM2007, pp. 187-195, 2007.
    3. Sheng Tang, YongDong Zhang, JinTao Li, Juan Cao, Huanbo Luan, Qiaoyan He, Xu Zhang. TRECVID 2007 Search Tasks by NUS-ICT, TREC Video Retrieval Evaluation Online Proceedings (TRECVID), 2007
    4. Xuefeng Pan, Jintao Li, Yongdong Zhang, Sheng Tang, Juan Cao, Retrieval Method for Same Video Content in Different Format based on Spatiotemporal Features¡±, 29th European Conference on Information Retrieval (ECIR), Rome, Italy, 2-5 April 2007.

    2006

    1. Juan Cao, Jintao Li, Yongdong Zhang, and Sheng Tang, A Novel Method for Spoken Text Feature Extraction in Semantic Video Retrieval, PCM06, pp. 270 ¨C 278, 2006