ACCV 2006 : Plenary Talks

Plenary Talks

Friday, January 13, 2006

Special Talk : Prof. Takeo Kanade, Carnegie Mellon University, Pittsburgh

16:00-17:00 Special Talk on "Factorization Methods in Computer Vision"

Saturday, January 14, 2006

Prof. Andrew Blake, Microsoft Research, Cambridge

08:30-10:00 Inauguration followed by Plenary Talk 1

Video Segmentation by Fusion of Colour, Contrast and Stereo

Technology advances mean that a stereo webcam could be manufactured and sold for essentially the same price as a monocular one. There are two outstanding advantages for teleconferencing in using stereo vision. First, automatic control of pan/tilt/zoom, which is possible monocularly, is particularly robust in stereo. Second privacy can be protected by obscuring background elements and replacing them with safer ones. For example a business conversation held at home could show only the talking head, against a bland video background, with inappropriate elements obscured. The first of these advantages is readily attainable and we describe progress towards achieving the second.

An algorithm will be described that is capable of real-time segmentation of foreground from background layers in stereo video sequences. Automatic separation of layers from colour/contrast or from stereo alone is known to be error-prone. Here, colour,contrast and stereo matching information are fused to infer layers accurately and efficiently. The "Layered Graph Cut" algorithm does not directly solve stereo. Instead it marginalises the stereo match likelihood over foreground and background hypotheses, and fuses it with a contrast-sensitive colour model that is learned on the fly. Segmentation is then solved efficiently and exactly by binary graph cut. The algorithm will be demonstrated in the application of background substitution and shown to give good quality composite video output.

Sunday, January 15, 2006

Prof. Andrew Zisserman, Oxford University

08:30-09:40 Plenary Talk 2

Category Recognition: Bags of Words and Beyond

There has been much recent research activity - and much recent success - in recognizing visual object categories (such as cars, faces, motorbikes) in images. The success has come from representing objects by sets of local iconic image patches, where each patch may be thought of as a "visual word" for describing part of the object. Surprisingly object categories can be recognized without including the spatial organization/location of the patches, and these models are referred to as a "bag of words" in analogy with similar models in the statistical text understanding literature.

In this talk I will describe two methods for learning bag of words models: an unsupervised approach where object category models are learnt from an unlabelled set of images; and a supervised approach where object category models are learnt from images obtained from Google image search. In both cases object category models are obtained by fitting with probabilistic Latent Semantic Analysis (pLSA), a model originally developed in the statistical text literature. The talk will conclude with some illustrations of how spatial organization can be added to these models.

Monday, January 16, 2006

Prof. Baba Vemuri, University of Florida

08:30-09:40 Plenary Talk 3

Information Theoretic Measures and Their Applications
to Computer Vision and Medical Imaging

Information theory has played a fundamental role in many fields of science and engineering including Computer Vision and Medical Imaging. In this talk, I will introduce various information theoretic measures that are used in achieving the goal of solving several important problems in Computer Vision and Medical Imaging, namely, image registration, point set registration, tensor field segmentation, image/shape retrieval etc. Recently introduced information theoretic measures such as, entropy defined on distributions, averages of Gaussian distributions (computed as the minimizer of the sum of squared J-divergences) and the well known Jensen-Shannon divergence etc. will be presented. I will show how each of these measures are used in solving the aforementioned application problems. The talk will be interspersed with several examples from each of the aforementioned applications.


For Authors Presentation Instructions Call for Papers Important Dates Instructions Paper Submission For Attendees Registration Travel Venue Accommodation Conference Program Plenary talks Tutorials General Information Advisory Committee Organizers Area Chairs Program Committee Sponsors Contact us	Plenary Talks Friday, January 13, 2006 Special Talk : Prof. Takeo Kanade, Carnegie Mellon University, Pittsburgh 16:00-17:00 Special Talk on "Factorization Methods in Computer Vision" Saturday, January 14, 2006 Prof. Andrew Blake, Microsoft Research, Cambridge 08:30-10:00 Inauguration followed by Plenary Talk 1 Video Segmentation by Fusion of Colour, Contrast and Stereo Technology advances mean that a stereo webcam could be manufactured and sold for essentially the same price as a monocular one. There are two outstanding advantages for teleconferencing in using stereo vision. First, automatic control of pan/tilt/zoom, which is possible monocularly, is particularly robust in stereo. Second privacy can be protected by obscuring background elements and replacing them with safer ones. For example a business conversation held at home could show only the talking head, against a bland video background, with inappropriate elements obscured. The first of these advantages is readily attainable and we describe progress towards achieving the second. An algorithm will be described that is capable of real-time segmentation of foreground from background layers in stereo video sequences. Automatic separation of layers from colour/contrast or from stereo alone is known to be error-prone. Here, colour,contrast and stereo matching information are fused to infer layers accurately and efficiently. The "Layered Graph Cut" algorithm does not directly solve stereo. Instead it marginalises the stereo match likelihood over foreground and background hypotheses, and fuses it with a contrast-sensitive colour model that is learned on the fly. Segmentation is then solved efficiently and exactly by binary graph cut. The algorithm will be demonstrated in the application of background substitution and shown to give good quality composite video output. Sunday, January 15, 2006 Prof. Andrew Zisserman, Oxford University 08:30-09:40 Plenary Talk 2 Category Recognition: Bags of Words and Beyond There has been much recent research activity - and much recent success - in recognizing visual object categories (such as cars, faces, motorbikes) in images. The success has come from representing objects by sets of local iconic image patches, where each patch may be thought of as a "visual word" for describing part of the object. Surprisingly object categories can be recognized without including the spatial organization/location of the patches, and these models are referred to as a "bag of words" in analogy with similar models in the statistical text understanding literature. In this talk I will describe two methods for learning bag of words models: an unsupervised approach where object category models are learnt from an unlabelled set of images; and a supervised approach where object category models are learnt from images obtained from Google image search. In both cases object category models are obtained by fitting with probabilistic Latent Semantic Analysis (pLSA), a model originally developed in the statistical text literature. The talk will conclude with some illustrations of how spatial organization can be added to these models. Monday, January 16, 2006 Prof. Baba Vemuri, University of Florida 08:30-09:40 Plenary Talk 3 Information Theoretic Measures and Their Applications to Computer Vision and Medical Imaging Information theory has played a fundamental role in many fields of science and engineering including Computer Vision and Medical Imaging. In this talk, I will introduce various information theoretic measures that are used in achieving the goal of solving several important problems in Computer Vision and Medical Imaging, namely, image registration, point set registration, tensor field segmentation, image/shape retrieval etc. Recently introduced information theoretic measures such as, entropy defined on distributions, averages of Gaussian distributions (computed as the minimizer of the sum of squared J-divergences) and the well known Jensen-Shannon divergence etc. will be presented. I will show how each of these measures are used in solving the aforementioned application problems. The talk will be interspersed with several examples from each of the aforementioned applications.
Organised by IIIT Hyderabad