Slides available here!

Speaker:

Michel Paindavoine, Global Sensing Technologies, France

Title:

Low-Power Neural Processor for Embedded Vision Applications

Abstract:

In order to realize autonomous vision systems (low-power consumption) with neuro-inspired algorithms inside, we studied in cooperation with CEA LIST a new architecture of a Neural Processor named PNeuro.

In this keynote we first describe vision neuro-inspired algorithms used nowadays. Then, we introduce in the second part the architecture of PNeuro processor. In the last part, we describe the implementation of the optimized Hmax and ConvNets neuro-inspired models using the new PNeuro accelarator in the context of face detection application. We show that PNeuro implemented in FDSOI 28nm technology can manage face detection from VGA images at 30 fps with less than 20mW. Same, PNeuro implemented onto FPGA working at 100MHz can detect faces from images with a 48x48 pixels resolution at 2,000 fps.

Bio:

Michel Paindavoine received the PHD in Electronics and Signal Processing from Montpellier University, in 1982 and the "Habilitation à Diriger les Recherches" from Burgundy University (Dijon, France), in 1991.

From 1982 to 1985, he was Research Engineer in industrial companies (Fairchild, Data Sud) and from 1985 to 1992, he was Assistant Professor at the Burgundy University.

He is Full Professor from 1992 in this university and is member of the LEAD laboratory (UMR 5022 CNRS & University of Burgundy) where he conducts researches in neuro-inspired electronics architectures. He is also co-founder in 2011 and CTO of GlobalSensing Technologies (GST) company specialized in embedded artificial network systems.

His main research topic concerns architectures for image acquisition and real time image processing & recognition. In this area, Professor Michel Paindavoine has published 150+ papers in international journals and conferences and has supervised 32 PhD students.



* If you wish to modify any information or update your photo, please contact the Publicity Chair at the following address.