Computer Science University

Timoth e Cour

e phone: 408-***-****

NEC Laboratories, 10080 N Wolfe Rd #Sw3350 Cupertino, CA 95014 ********.****@*****.***

Visa Status: US Green Card Holder http://www.timotheecour.com

Research Interests

Computer Vision: object detection, recognition, segmentation, image retrieval, gesture recognition.

Machine Learning: weakly supervised learning, boosting, large scale learning.

Education

University of Pennsylvania Philadelphia, PA, USA

Ph.D. in Computer and Information Science (Advisor: Ben Taskar) 2009

Weakly supervised learning from multiple modalities: exploiting video, audio& text for video understanding

Committee: Kostas Daniilidis, Fernando Pereira, Camillo J. Taylor, Andrew Zisserman

M.S. in Computer and Information Science (Advisor: Jianbo Shi) 2005

Ecole Polytechnique (the best rated French engineering school) Paris, France

B.S. Applied Mathematics and Computer Science 2003

Research Fellowship from University of Pennsylvania and from the Ecole Polytechnique Foundation

Carnot Foundation Fellowship, 2003 (3 awardees from Ecole Polytechnique each year)

Research and Work Experience

NEC Labs Cupertino, CA, USA

Research Scientist, Media Analytics department, with Kai Yu and Yuanqing Lin. since 2010

Real time pedestrian, face, and hand detection, large scale image retrieval with state-of-the-art results on 4

public datasets, gesture recognition from webcam or kinect sensor, 3D reconstruction.

Member of the winning team at the ImageNet Challenge, 2010 (1K categories, 1M images).

Demos at the NEC Open House in Tokyo in 2010 (image retrieval) and 2011 (gesture based mouse control).

INRIA / Ecole Normale Sup rieure - Willow project

e Paris, France

Postdoctoral researcher with Jean Ponce and Francis Bach. 2009-2010

Learning from weakly annotated images, multiple instance boosting, video alignment using plot summaries.

GRASP Lab, University of Pennsylvania Philadelphia, PA, USA

Graduate Research Fellow with Ben Taskar and Jianbo Shi 2003-2009

Weakly-supervised learning, object detection& segmentation, graph matching, approximate inference.

Microsoft Research Redmond, WA, USA

Research Internship with Paul Viola and Michael Shilman summer 2005

Recognition of handwritten mathematical expressions using e cient two-dimensional parsing.

Intern Supervision

Kevin Lai (University of Washington, 2012), Olivier Duchenne (ENS, 2011), Jiangping Wang (UIUC, 2011),

Ugo Jardonnet (ENS, 2010), Remi Cuingnet (Ecole Polytechnique, 2006), Pierre Fournier (Ecole Polytechnique,

2005), Florence Benezit (Ecole Polytechnique, 2004), Nicolas Gogin (Ecole Polytechnique, 2004).

Teaching

University of Pennsylvania: Teaching Assistant: Computer Vision (CSE 399), Mathematical Foundations of

Computer Science (CSE 260), Automata, Computability, and Complexity (CSE 262). 2004 - 2006

Ecole Polytechnique: Tutoring in Maths and Physics for undergraduate students. 2000-2003

Invited talks

Talking Pictures: Temporal Grouping and Dialog-Supervised Person Recognition. Oxford

Brookes/KTH/VGG/Willow Workshop, 2010.

Weakly Supervised Learning for Video Understanding and Object Recognition. NEC Labs,

Cupertino, 2010; Intel Labs, Seattle, 2010; Google, Mountain View, 2010, MSR-INRIA Workshop on Computer

Vision and Machine Learning, 2010.

Talking Pictures: Temporal Grouping and Dialog-Supervised Person Recognition in Video. CVPR

2009 Workshop on Visual and Contextual Learning from Annotated Images and Videos.

Movie/Script: Alignment and Parsing of Video and Text Transcription. Google, Mountain View, 2008.

Object segmentation and recognition in image datasets and movies. ENS, Paris, 2007.

Video Deconstruction: Revealing narrative structure through image and text alignment. NIPS

2007 Workshop on the grammar of vision.

Publications

In Computer Vision and Machine Learning, conferences such as CVPR, ICCV, ECCV, NIPS, AISTATS are

highly refereed. Google scholar: http://scholar.google.com/citations?user=pkFzb9QAAAAJ

X. Wang, M. Yang, T. Cour, S. Zhu, K. Yu Exploring E cient Contextual Weighting for Image Retrieval using

Vocabulary Trees. Submitted to IJCV 2012.

S. Zhang, M. Yang, T. Cour, K. Yu Query Speci c Fusion for Image Retrieval. ECCV 2012.

X. Wang, M. Yang, T. Cour, S. Zhu, K. Yu, and T. X. Han. Contextual Weighting for Vocabulary Tree based

Image Retrieval. ICCV 2011.

T. Cour, B. Sapp, B. Taskar. Learning from Partial Labels. JMLR 2011.

Y. Lin, F. Lv, S. Zhu, M. Yang, T. Cour, K. Yu, L. Cao, T. Huang. Large-scale image classi cation: fast

feature extraction and SVM training. CVPR 2011. (16 citations)

T. Cour, B. Sapp, A. Nagle, B. Taskar. Talking Pictures: Temporal Grouping and Dialog-Supervised Person

Recognition. CVPR 2010.

T. Cour Weakly Supervised Learning from Multiple Modalities: Exploiting Video, Audio and Text for Video

Understanding. PhD Thesis, University of Pennsylvania 2009.

T. Cour, B. Sapp, C. Jordan, B. Taskar. Learning from Ambiguously Labeled Images. CVPR 2009.

(34 citations)

T. Cour, C. Jordan, E. Miltsakaki, B. Taskar. Movie/Script: Alignment and Parsing of Video and Text

Transcription. ECCV 2008. (33 citations)

T. Cour, J. Shi. Recognizing objects by piecing together the Segmentation Puzzle. CVPR 2007.

(49 citations)

T. Cour, J. Shi. Solving Markov Random Fields with Spectral Relaxation. AISTATS 2007. (24 citations)

T. Cour, P. Srinivasan, J. Shi. Balanced Graph Matching. NIPS 2006. (77 citations)

T. Cour, F. Benezit, J. Shi. Spectral Segmentation with Multiscale Graph Decomposition. CVPR 2005.

(230 citations)

T. Cour, N. Gogin, J. Shi. Learning spectral graph segmentation. AISTATS 2005. (17 citations)

Technical reports or Workshop papers:

T. Cour Convex Relaxations for Markov Random Field MAP estimation. University of Pennsylvania,

Philadelphia 2008.

T. Cour, B. Taskar. Video Deconstruction: Revealing narrative structure through image and text alignment.

Workshop on the Grammar of Vision, NIPS 2007.

Professional Activities

Member of IEEE (2006 - present).

Best Reviewer award, CVPR 2011.

Program committee / reviewer (conferences): ICCV 2007, 2009; ECCV 2010; CVPR 2009, 2010, 2011,

2012; NIPS 2009, 2010, 2011; ICML 2010, 2012; AAAI 2007; NESCAI 2007; IJCAI, 2009.

Reviewer (journals): PAMI 2007, 2008, 2008, 2009, 2010; JMLR 2010, 2011; CVIU 2007, 2008, 2010; IEEE

Transactions on Image Processing 2007; IEEE Transactions on Information Theory 2007.

Open-source software

Open source software in Matlab, C++, available at http://www.timotheecour.com.

Convex Learning from Partial Labels Toolbox: Implementation of the partially labeled multiclass

classi cation algorithm introduced in Learning from Partial Labels, JMLR 2011. This also includes the

Annotated Faces on TV Dataset: faces extracted, aligned and annotated from 8 TV shows.

Graph matching toolbox: State of the art and scalable subgraph matching algorithm. Uses spectral

relaxation with a ne constraints and bistochastic normalization. 5000+ page views since 2010.

Normalized Cuts Segmentation Code: Image segmentation and data clustering using Normalized Cuts.

15000+ page views since 2010.

Multiscale Normalized Cuts Image segmentation toolbox: Linear time image segmentation based on

multiscale graph decomposition. Extensively tested and downloaded. 13000+ page views since 2010.

Technical skills

Programming languages: C++, C, D, Python, Matlab, Java, C#.

Libraries: opencv, opengl, openni, libfreenect, phobos, stl, boost.

Tools: Eclipse, Xcode, Visual Studio, SWIG, LaTeX, gdb, hadoop, git, svn, waf, fbuild.

Operating systems: Mac-OSX, Linux, Windows.

References

Prof. Ben Taskar, University of Pennsylvania, ******@***.*****.***

Prof. Kostas Daniilidis, University of Pennsylvania, ******@***.*****.***

Prof. Andrew Zisserman, Oxford University, **@******.**.**.**

Prof. Jean Ponce, Ecole Normale Sup rieure, ****.*****@***.**

e

Prof. Michael Shilman, Seoul National University, *******@*******.***

Contact this candidate