Plenary Talk Abstracts
K.S. Fu Prize Lecture:
Dictionaries, Manifolds and Domain Adaptation Methods:
New Solutions to Old Problems in Pattern Recognition
Rama Chellappa, IAPR Fellow (USA)
Abstract: Feature extraction or representation of patterns and adaptation of classifiers designed using training data to be effective on testing data are two fundamental problems in pattern recognition. In this talk, I will discuss new solutions to these problems based on theories of dictionary learning, analytic manifolds and domain adaptation with applications in image and video-based recognition. Specifically, I will discuss methods for representing images and videos using linear and non-linear dictionaries and analytical manifolds. I will then discuss methods for adapting the dictionaries and manifold representations for addressing shifts in data distributions due to changes in pose, illuminations, spatio-temporal sampling and blur with applications in recognition of faces, expressions, objects and actions.
J. K. Aggarwal Prize Lecture:
Generalized Principal Component Analysis (GPCA)
and Sparse Subspace Clustering (SSC)
René Vidal (USA)
Abstract: In the era of data deluge, the development of methods for discovering structure in high-dimensional data is becoming increasingly important. Traditional approaches often assume that the data is sampled from a single low-dimensional manifold. However, in many applications in signal/image processing, machine learning and computer vision, data in multiple classes lie in multiple low-dimensional subspaces of a high-dimensional ambient space. In this talk, I will present methods from algebraic geometry, sparse representation theory and rank minimization for clustering and classification of data in multiple low-dimensional subspaces. I will show how these methods can be extended to handle noise, outliers as well as missing data. I will also present applications of these methods to video segmentation and face clustering.
Takeo Kanade (Japan)
Abstract: For understanding the behavior, intent, and environment of a person, the surveillance metaphor is traditional; that is, install cameras in the environment and observe her and her interaction with other people and environment from them. Instead, we argue that the First-Person Vision that senses the environment and her activities from her point of view is more advantageous with available images about her environment from her own view points and with readily available information about her head motion and gaze. We have been working in this paradigm for a while, and this talk will present the current progresses in the First Person Vision - the ideas, devices, algorithms, and example applications.