GPI Seminar Series: José M. Álvarez

José M. Álvarez

José M. Álvarez, computer vision researcher at NICTA, Australia
Compacting ConvNets for end to end Learning
Tuesday February 2nd, 16h, MERIT Room D5-010

Abstract:
Convolutional neural networks have achieved considerable success in many tasks in computer vision such as image classification, object detection / recognition or semantic segmentation. These networks are computationally demanding and not always feasible for embedded platforms where power and computational resources are relevant. Recent works have shown significant redundancy in the parameters of these networks. This over parametrization seems necessary to overcome the challenges existing in highly non-convex optimization problems. In this talk I review recent techniques to speed up and reduce the parameter redundancy existing in current networks.

Short Bio:
Jose M. Alvarez is a computer vision researcher at NICTA (Australia). His main research is focused on data-driven methods for large-scale dynamic scene understanding. In particular he is interested in deep learning and efficient non-parametric methods for embedded platforms working in real-world environments.
Jose M. Alvarez received his PhD degree from the Universitat Autònoma de Barcelona (UAB) in October 2010, under the supervision of Dr. Antonio M. López and Prof. Theo Gevers, with a thesis entitled “Combining Context and Appearance for Road Detection”. This thesis received the best PhD thesis award from the Computer Science Department at the UAB. During his PhD program, he visited the ISLA group at the University of Amsterdam (in 2008 and 2009) under the supervision of Prof. Theo Gevers, and the Group Research Electronics at Volkswagen (in 2010) under the supervision of Dr. Thorsten Graf. During that time he also worked as an engineering in several industrial projects developing a headlight evaluation system for SEAT (patented) under the supervision of Dr. Felipe Lumbreras and Dr. Joan Serrat. As a freelance, he also developed different systems such as an on-board multi-modal acquisition platform, or the Volumeter, a volume measurement system for uncovered loads in vehicles (patented). Subsequently, he worked as a post-doctoral researcher at the Courant Institute of Mathematical Science, New York University, with Professor Yann LeCun. His main research topic was convolutional neural networks for road scene understanding.