Talk WHAT: M. Alex O. Vasilescu (University of Toronto, Department of Computer Science; New York University, Courant Institute of Mathematical Sciences): "A Tensor Framework for Modeling the Perceptual Signatures of People" WHEN: Monday, April 11, 2005, 4:00PM EST WHERE: Bartos Theatre, MIT Media Lab (E15) SUMMARY: An individual can be recognized from several perceptual modalities—such as facial appearance, gestures, gait, or voice—under varying conditions. For example, we recognize faces even when the viewpoint, lighting, or facial expression changes, and we can easily recognize individuals depicted by a skilled impersonator mimicking their movements. This suggests that our brains compute invariant perceptual representations of people, which enable recognition under unconstrained conditions. Vasilescu's research aims to imbue computers with learned generative models that, among many other uses, support this human-like capacity for recognition. To this end, she will present a tensor algebraic model that learns (nonlinear) manifold representations of facial image ensembles including different people, expressions, viewpoints, and illuminations. This multilinear model, called TensorFaces, supports a potent new machine-learning approach to automated face recognition. It is computed via a tensor decomposition, known as the N-mode SVD, which is an extension to tensors of the conventional matrix singular value decomposition (SVD) associated with (linear) principal and independent components analysis (PCA/ICA). Vasilescu will also summarize the application of tensor algebra to human motion capture data, leading to decompositions that we call "human motion signatures," which are useful for action recognition and motion synthesis. Our tensor algebraic framework is also valuable in image synthesis. She will demonstrate TensorTextures, a multilinear approach to image-based rendering, which learns the interaction between viewpoint, illumination, and geometry that determines the appearance of complex surfaces. BIO: M. Alex O. Vasilescu was educated at MIT and the University of Toronto, where she is currently a PhD candidate under the supervision of Prof. Demetri Terzopoulos. She is also a research scientist at New York University's Courant Institute of Mathematical Sciences. She has done research at the MIT Artificial Intelligence Lab and has interned at research centers of IBM, Intel, Compaq, and Schlumberger corporations. She has published research papers in computer vision and computer graphics, particularly in the areas of face recognition, human motion analysis/synthesis, image-based rendering, and physics-based modeling (deformable models). She has given several invited talks about her work and has three patents pending. Her face recognition research, known as TensorFaces, is being funded by the TSWG, the Department of Defense's Combating Terrorism Support Program. She has been named by MIT's Technology Review Magazine to their 2003 TR100 List of Top Young Innovators. MIT Media Laboratory Home Page | Events Main Index