Subspace2007

Subspace 2007

Workshop in conjunction with ACCV2007

Tokyo, Japan

Nov. 19 , 2007

Program (10/16)

Proceedings

Submissions The site is opened. (7/24)

Publishes

CFP

ACCV2007

Workshop goals:

Subspace methods such as CLAFIC [1] are not only important theory for solving many pattern recognition problems in computer vision, but also they have been widely used as a practical methodology for a wide variety of real applications. In particular, subspace methods have been studied intensively in the field of character recognition, contributing to a number of commercial optical character recognition (OCR) systems. Although the firstly proposed subspace method is nearly three decades old, a number of new techniques based on the same approach are still proposed every year in emerging fields between computer vision and other related technologies. For example, it is interesting to see that the Mutual Subspace Method, which is one of the most successful variants of subspace methods, is indeed almost identical to the Canonical Correlation Method in multivariate statistical analysis. The concept of subspace methods is also strongly related to the factorization method. Unfortunately, however, the significance of subspace methods is not yet fully recognized in the community of computer vision, despite the notable success of commercial products based on the subspace method. The goal of this workshop is to share the prominent potential of subspace methods with researchers working on various problems in computer vision, and encourage interactions leading to further developments of subspace methods. The fundamental theories of subspace methods and their applications in computer vision will be discussed at the workshop. Prior to this workshop, we successfully organized a domestic workshop focused on subspace methods called Subspace2006 last summer. The website is found at [2]. We consider Subspace2007 as an international version extended from Subspace2006. We believe that Subspace2007 will stimulate fruitful discussions among participants and provide them novel ideas for future research in computer vision.

[1] S. Watanabe, N. Pakvasa, Subspace method in pattern recognition, in Proc. 1st Int. J. Conf on Pattern Recognition, Washington DC, 1973.

[2] http://www.viplab.is.tsukuba.ac.jp/~ss2006/index.html

Organizers

Roberto Cipolla (University of Cambridge),

Kazuhiro Fukui (University of Tsukuba),

Hitoshi Sakano (NTT Data),

David Suter (Monash University)

Program committee:

Sang-Woon Kim (Myongi University),

Xi Li (University of Tsukuba),

Ken-ichi Maeda (Toshiba),

Atsuto Maki (Kyoto University),

Hiroshi Murase (Nagoya University),

Shree K. Nayar (Columbia University),

Erkki Oja (Helsinki University of Technology),

Shinichiro Omachi (Tohoku University),

Seiichi Uchida (Kyushu University),

Shin'ichi Satoh (National Institute of Informatics),

Atsushi Sato (NEC),

Yoichi Sato (University of Tokyo)

Liwei Wang (Peking University)

Scope:

The topics of interest include, but are not limited to, the following:

- Theoretical foundations of subspace methods

Iijima equation, Watanabe's formulation of CLAFIC theory,

Subspace method as degenerated Gaussian, geometry for subspace methods, Illumination cone

- Theoretical extensions of subspace methods

Construction methodology of subspace (PCA, CCA, FA, ICA, etc),

nonlinear subspace method (Gnanadesikan's method, principal manifold, MLP based method, kernel PCA and etc.), Mutual Subspace Method and their theoretical extensions, similarity measure with subspace

- Related technology

Mathematical feature extraction (n-mode SVD), eigenspace method, discriminant analysis, factorization method

- Applications

Object recognition, face recognition, gesture recognition, character recognition,　motion analysis, scene analysis, robot vision, biometrics, anomaly (unusual) detection，data visualization and other novel applications

Important dates: (Extended on 7/17)

Deadline of submission	August 31st, 2007
Notification to authors	September 15th, 2007
Submission of camera ready	September 30th, 2007

Registration:

It is requested that all participants of the workshop including speakers register ACCV'07. On-line registration of workshops, tutorials and the conference will be available via ACCV web page. Registration site will open shortly.

Submission information:

SS2007 invites submissions of original, previously unpublished work, and also welcomes, with some restrictions, submissions which are closely related to work submitted to ACCV2007. An SS2007 paper should differ materially from ACCV2007 paper(s) in the following ways:

· Recent results too late for the ACCV2007 deadline

· Motivations and broader implications of work presented in more detail than in your ACCV paper(s)

Instructions for presenters:

Oral presentations:

Oral presentations will be held in the convention hallon on the 2nd floor of the A-block (General Research Experiment Building. See also "2. Map of Komaba Research Campus" in the map page for detail.): it is equipped with a LCD projector, microphones and laser pointers. For presentations, you may use your own computer: please make sure to test its video output beforehand and be familiar with its configuration.

Each oral presentation is allocated a total of 20 minutes: please ensure that you provide a sufficient question period at the end of your talk. You should aim for 15-17 minutes of presentation, followed by 3-5 minutes for questions. The Session Chair is responsible for maintaining the pace of presentations, and will signal the end of your allocated time.

For all the presenters:

All the presenters including the oral presenters are expected to have poster presentations

Please prepare your poster presentation.

Poster presentations:

Poster presentations will be held in the Foyer of the Convention Hall on the 2nd floor in A-block (General Research Experiment Building. See also "2. Map of Komaba Research Campus" in the map page for detail.). The poster session is co-operated with the workshop of Multi-dimensional and Multi-view Image Processing, (General Chair: R. Taniguchi, Program Chair: R. Kurazume). You can see all the posters in two workshops at once! You will be able to set up your poster from 9:00 on Monday, November 19 (30 minutes prior to the beginning of the oral session). You will find the location of your poster panel from the number in the Timetable above. The poster board surface allocated to each author consists of two adjacent panels, each measuring approximately 120cm high x 90cm wide. Therefore, you can put up two A0 posters (119cm x 84cm) or one B0 (103cm x 146cm) poster on the board. Conference staff will provide you with the proper pins to attach your poster. Authors are expected to be at their poster during the poster session.