Seminar by Dr. Tomasz Kryjak (AGH University of Krakow, Poland) "Embedded event-based-vision" (introduced by Frédéric Pétrot - SLS team)

Abstract: The use of SoC FPGAs (System on Chip Field Programmable Gate Arrays) to accelerate event data processing could unlock the full potential of event cameras in advanced robotic domains where fast processing, low latency and energy efficiency are critical factors. We will start this talk with an overview of the state of the art based on 60 research papers published in the years 2012-2024. Then, we will present our own research on the following topics: event data filtering, object detection using convolutional networks (CNNs) with logarithmic (power of two) quantisation, and object classification using graph convolutional networks (GCNs). In the last part of the talk, other research activities of the Embedded Vision Systems Group in the area of embedded vision, preception and control will be very briefly presented.

Bio: Tomasz Kryjak is an Assistant Professor at the Embedded Vision Systems Group, Department of Automatic Control and Robotics, AGH University of Krakow, Poland. His research focuses on embedded perception and control systems implemented in SoC FPGAs (using hardware/software co-design). His application area includes mobile robotics (drones and self-driving cars), advanced driver assistance systems (ADAS), and advanced video surveillance systems (object detection, recognition and tracking). He works with event cameras, embedded AI, and neuromorphic computing. He is an IEEE Senior Member, a steering committee member of the DASIP and DSD conferences, a TPC member of the ARC conference and an associate editor of the Microprocessors and Microsystems Journal. He has authored and co-authored more than 120 scientific papers.