Daniel Huttenlocher
John P. and Rilla Neafsey Professor

of Computing, Information Science and Business
and Stephen H. Weiss Fellow


Dean of Computing and Information Science

Professor, Computer Science Department

Professor, Johnson Graduate School of Management

Administrative: cis-dean “at” cs.cornell.edu
Academic: dph “at” cs.cornell.edu
607 255-9188

Research    Selected Papers    Recent Talks    Teaching

Research

My research in computer vision ranges from theoretical algorithms (using techniques from computational geometry and graph algorithms) to the development of end-to-end systems that apply visual matching and recognition techniques.  Some of my computer vision work includes:

  • Image matching and comparison
    • Graphical models for object recognition – a Bayesian approach to object recognition that takes into account local spatial dependencies among features (implementations are available for several papers, see below)
    • Pictorial structures for recognizing generic objects composed of multiple parts, such as people and faces
    • Hausdorff-based methods for visual matching and recognition (an old C implementation for SunOS is also available)
    • Performance evaluation and modeling of recognition methods
  • Object tracking and video monitoring
    • Object tracking and identification in video
    • Video surveillance and monitoring for detecting vehicles in built-up areas
    • Pictorial structure models for tracking people
  • Efficient algorithms for low-level vision

 

My research on the Web and large-scale social networks is focused on developing models and measures that allow us to better study and understand how people interact with one another, particularly in computer-mediated environments.  For instance, how does the structure of a social network influence one’s propensity to undertake certain actions?  There is a long history of study of such questions in the social sciences, primarily for small-scale networks that can be mapped out by hand.  While computer and information scientists have been studying large-scale networks, their focus has been more on the network properties and less on testing and extending existing social science theories of social interaction.

 

My work on autonomous vehicles grows out of my role as co-leader of Team Cornell’s entry in the DARPA Urban Challenge race.  Our vehicle was one of 6 out of 11 finalists (and 35 semi-finalists) to complete the race.  The students on our team made all the design decisions and did an outstanding job overall.  While the Urban Challenge and the Grand Challenges before that have led to enormous progress in autonomous driving, the race also highlighted some fundamental research questions that remain to be addressed in order to enable perception and reasoning about the actions and intentions of other vehicles.  For example, the fender bender between our vehicle and MIT’s could have been avoided if either system had been able to perceive what the other was doing over an extended time period (i.e., perceive actions in addition to locations and velocities). Such issues form the basis of my current and planned research in the area.

 

My research on geometric algorithms includes efficient algorithms for computing Hausdorff distances and related distance transforms, as well as techniques for comparing three-dimensional protein structures.

 

My work on interactive document systems has often incorporated computer vision techniques, and includes:

  • DigiPaper: a highly compact, universally viewable document image format (now the Silx project at PARC)
  • CoNote: a system for supporting collaboration with shared documents
  • Automatically constructing browse-able presentations from a video recording of a lecture

 

My interest in electronic trading systems focuses primarily on illiquid or thinly-traded markets, where conventional auction and exchange mechanisms are not very effective ways of making trades.

 

 

My interest in software development methodologies stems from my involvement in the creation of large, complex software systems at Xerox Corporation and Intelligent Markets.  Through these activities I have come to believe that:

·         We need more clearly stated principles of software development that can help guide the choice of appropriate development practices for a given software project.  Too much of software development methodology is aimed at practices without a deeper understanding of underlying principles.

·         Training of software developers requires a better understanding of the fact that software is an intellectual work product, in many ways more akin to a legal brief, an architectural plan, or an ad campaign, than to a physical or electronic device.  As such, I believe that training programs for software developers should be modeled more on professional programs such as law, business, architecture or design.


 

 

Selected Papers

Object Recognition and Detection

Segmentation and Low-Level Vision Algorithms

·         Generating Sharp Panoramas from Motion-blurred Videos, Proceedings CVPR, 2010 (with Y. Li, S.B. Kang, N. Joshi and S. Seitz).

·         Learning for Optical Flow using Stochastic Optimization, Proceedings of European Conference on Computer Vision (ECCV), 2008 (with Y. Li).

·         Sparse Long-Range Random Field and its Application to Image Denoising, Proceedings of European Conference on Computer Vision (ECCV), 2008 (with Y. Li).

·         Learning for Stereo Vision Using the Structured Support Vector Machine, Proceedings of the IEEE Computer Vision and Pattern Recognition Conference, 2008 (with Y. Li).

·         Efficient Belief Propagation with Learned Higher-Order Markov Random Fields, Proceedings of ECCV, 2006 (with X. Lan, S. Roth and M. Black).  

·         Efficient Belief Propagation for Early Vision, to appear in Intl. Journal of Computer Vision (with P. Felzenszwalb).  Conference version from IEEE CVPR, Vol 1, pp. 261-268, 2004.   CODE

·         Efficient Graph-Based Image Segmentation, Intl. Journal of Computer Vision, vol. 59, no. 2, pp. 167-181, 2004 (with P. Felzenszwalb).    CODE

The Web and Large-Scale Social Networks

·         Governance in Social Media: A case study of the Wikipedia promotion process, Proceedings of AAAI International Conference on Weblogs and Social Media (ICWSM) 2010 (with J. Leskovec and J. Kleinberg).

·         Predicting Positive and Negative Links in Online Social Networks, Proceedings of Nineteenth International World Wide Web Conference (WWW) 2010 (with J. Leskovec and J. Kleinberg).

·         Signed Networks in Social Media, Proceedings of ACM CHI 2010 (with J. Leskovec and J. Kleinberg).

·         Mapping the World's Photos, Proceedings of Eighteenth International World Wide Web Conference (WWW) 2009 (with D. Crandall, L. Backstrom and J. Kleinberg).

·         Feedback Effects between Similarity and Social Influence in Online Communities, Proceedings of Fourteenth ACM Conference on Knowledge Discovery and Data Mining (KDD) 2008 (with D. Crandall, D. Cosley, J. Kleinberg and S. Suri).

·         Group Formation in Large Social Networks: Membership, Growth, and Evolution, Proceedings of Twelfth ACM Conference on Knowledge Discovery and Data Mining KDD (with L. Backstrom, J. Kleinberg and X. Lan), 2006. 

·         Traffic-Based Feedback on the Web. Proceedings of the National Academy of Sciences, 6 January 2004 (with J. Aizen, J. Kleinberg and T. Novak).

·         Fast Algorithms for Large State Space HMM’s with Applications to Web Usage Analysis, Advances in Neural Information Processing Systems (NIPS) 16, December 2003 (with P. Felzenszwalb and J. Kleinberg).   

Tracking

Geometric Algorithms

Interactive Document Systems

 

Selected Talks

·         Structured Models in Computer Vision, Workshop on Structured Models at CVPR 2010

·         Social Data, HICSS-43 Keynote talk, 2010

·         Tutorial on Pictorial Structures, ICVSS 2009

·         Mapping the World’s Photos: Communal Perception, ETHZ, EPFL, ENS 2009

·         Team Cornell’s Entry in DARPA Urban Challenge, 2007-08 (video of Google Techtalk)

·         Object Recognition Without Feature Detection, Oxford 2007.

·         Speeding Up Belief Propagation for Early Vision, MSRI Workshop, Feburary 2005

Teaching

In the past I have taught CS3110, functional programming and data structures, NBA6000 on the strategic role of information technology and NBA6100 on the management of technology-driven businesses using bio, info and nano technologies (jointly with Bruce Ganem).   I last taught computer vision in Spring 2008, CS664.

Professional Activities

For many years (1988-1999) I worked with Xerox PARC on electronic document image processing.  More recently I have been working with Intelligent Markets on electronic trading systems.

In 1998-99 I chaired the Cornell Task Force on Computing and Information, which led to the creation of the Faculty of Computing and Information Science.  In 2005-06 I also chaired the Cornell Task Force on Wisdom in the Age of Information.

I was general co-chair for CVPR 2009 in Miami and CVPR in 2006 in NYC as well as program co-chair of CVPR  in 2001 and 1997 (CVPR is the main North American computer vision conference).

I serve on the Board of Directors of the MacArthur Foundation, which has reinforced my view of the important role for information technology in achieving social good.

Other Interests

My favorite non-computer-geek activity is snowboarding (but without the mtv-extreme-sports-way-too-cool attitude).

 

Last Updated: August 2010

Back to the CS Dept. Home Page