DASIP 2016 : Special Sessions

22 Mars 2016

---

Catégorie : Conférence internationale

---

DASIP provides an inspiring international forum for latest innovations and developments in the field of leading edge embedded signal processing systems. The conference features Special Sessions with the purpose of introducing the DASIP community to relevant hot topics that were not covered by previous editions of the conference. Five Special Sessions will be opened for the 2016 edition of DASIP.

More informations are available at: http://ecsi.org/dasip

Abstract submission deadline: April 22, 2016

Paper submission deadline: April 29, 2016

Special Session 1 : MEthods and TOols for Dataflow PrOgramming

Session Chairs:

• Johan Lilius, Abo Akademy, Finland, johan.lilius@abo.fi
• Maxime Pelcat, IETR, France, Maxime.Pelcat@insa-rennes.fr

Abstract: The dataflow model of computation proposes a powerful approach to describe parallel computations by characterising them through their dependencies. This special session addresses the use of dataflow models in the context of embedded systems. Topics of particular interest are semantics, Programming languages and compilers, Models of Signal and Image processing algorithms and applications, Model classification, Model to model transformations, Automatic code generation, Algorithms and applications implementations, and run-time management of Dataflow computations.

Special Session 2 : Advances in Heterogeneous Reconfigurable Multicore Signal Processing Platforms

Session Chairs

• Jari Nurmi, Tampere University of Technology, Finland, jari.nurmi@tut.fi
• Diana Göhringer, Ruhr University Bochum, Germany, diana.goehringer@ruhr-uni-bochum.de
• Waqar Hussain, Tampere University of Technology, Finland, waqar.hussain@tut.fi

Abstract: The emergence of latest wireless standards and the upcoming realization of Internet-of-Things (IoT) scenario will require a tremendous amount of intensive signal processing. The traditional systems do not provide sufficient resources to meet the new criteria and on the other hand, multicore scaling is obstructed by the utilization-wall. In such a situation, heterogeneity and reconfigurability play an important role. These platforms can be reconfigured dynamically to the requirements of a set of applications and can be tailored to custom performance needs. The custom design and reconfiguration features of these platforms can save power and energy dissipation to unprecedented levels. This special session aims to bring together worldwide state-of-the-art research on reconfigurable multicore and manycore platforms from both academia and industry, to address the challenges by proposing intelligent solutions.

Special Session 3 : Computer Vision and Image Analysis

Session Chairs:

• Rachel Ababou, St Cyr Coetquidan, France, rachel.ababou@st-cyr.terre-net.defense.gouv.fr
• Patrice Delmas, Auckland University, New Zealand, p.delmas@auckland.ac.nz

Abstract: Computer vision and Image analysis have benefited of five decades of intense development. The complexity of the algorithms in this field can now benefit from dramatically increasing computational power to offer viable applications going from embedded systems to cloud computing. The main challenge is now to incorporate practical computer vision and image analysis in real life applications. This special session will accept submissions on all aspects of Computer Vision, Image Processing, Visualisation, Virtual and Artificial Reality, Artificial Life, Embedded Systems and Robotics with an emphasis on practical applications involving 3D vision at large.

Special Session 4 : Automotive Parallel Computing Challenges – Architectures, Applications, and Tricks

Session Chairs

• Walter Stechele, Technical University of Munich, Germany, Walter.Stechele@tum.de

• Tomasz Kryjak, AGH University of Science and Technology, Poland, tomasz.kryjak@agh.edu.pl
• Lionel Lacassagne, University Pierre and Marie Curie, Paris 6, France, lionel.lacassagne@lip6.fr
• Dominique Houzet, Grenoble University, France, dominique.houzet@gipsa-lab.grenoble-inp.fr 

• Martin Danek, daiteq, Czech Republic, martin@daiteq.com

Abstract: Many of today’s cars provide driver assistance functionality implemented on a distributed network of multicore processing units, including distance control, lane assist, traffic sign and pedestrian detection. Autonomous vehicles are about to appear on public streets, supporting the trend of bringing more functionality from autonomous driving into standard vehicles.

The focus of this special session will be on computational challenges and solutions related to automotive parallel computing. Approaches from outside of automotive, i.e. from a more general class of real-time vision applications with information fusion, which might be adopted in automotive solutions in the future, are welcome.

Topics include design space exploration and run time management of embedded heterogeneous MPSoC, including GPU and FPGA, scheduling and memory management for mixed criticality applications, vision, lidar and radar processing, sensor data fusion, machine learning approaches, safety and security aspects. Design goals include high computing performance at low power and low cost, as well as various optimization tricks to reduce data bandwidth, amount of information stored and/or processed while maintaining reliability of the results (convergence, error margin).

Special Session 5 : Neuro-inspired algorithms and architectures

Session Chairs:

• Michel Paindavoine, Université de Bourgogne, France, michel.paindavoine@u-bourgogne.fr
• Nicolas Farrugia, Telecom-Bretagne, France, nicolas.farrugia@telecom-bretagne.eu

Abstract: The pinnacle of evolution in complexity of biological systems is arguably the human brain, which enabled the rise of societies, art and technologies. While we are still far from understanding the brain, neuroscientific research has begun to shed light on the way complex perceptual issues or intelligent behavior are implemented by neural systems. As a consequence, innovative artificial architectures are emerging, and take advantage of such advances in neurosciences by implementing neuro- inspired algorithms. Interestingly, by definition, neuro-inspired algorithms are tightly linked with an actual hardware implementation, and most likely an efficient one in terms of functionality, robustness or energy. Therefore, the joint study of neuro-inspired algorithms and architectures holds a great potential for a better understanding of efficient hardware design, e.g. in terms of computational speed, resource usage, power consumption.

In this special session, we welcome submissions pertaining to algorithms and architectures inspired from biological neural systems (human or other species). Applications include, but are not limited to, efficient processing, classification or interpretation of sensory information, memory systems, reasoning or artificial intelligence. We encourage submissions that describe actual hardware implementations (FPGA / ASIC or GPU), however we will also consider neuro-inspired architecture models and design methods.