GPU4EO Challenge 2015 is an international initiative which involves students, researchers and professionals in a challenge aimed at improving the performance of remotely sensed data processing using the capacity of graphics processing units (GPUs). Teams are asked to use and process Earth observation satellite data with NVIDIA k40 GPU and DORIS, an open source software package, to obtain the best performance, as determined by the fastest processing time.
The challenge is open to teams composed of students and researchers belonging to research institutions and/or professionals. Teams may consist of
- Researchers coming from research centers and/or students and professionals who operate outside of the Puglia, Italy, region
- Researchers coming from public research centers operating in the territory of the Puglia region
- Teams of students and / or professionals coming from Puglia or studying at the Universities of Puglia.
A prize of €5.000, as well as other honors, will be awarded to winning teams: € 3,000 to the first ranked team, and € 2,000 to the first ranked team operating in Puglia. The award ceremony will be preceded by a technical workshop in which participants in the Challenge will present the team’s work and will illustrate activities and achievements. The committee also will cooperate in submitting a scientific paper in which the results will be illustrated by the winning team.
Challenge objectives
The goal of GPU4EO Challenge 2015 is “to stimulate the adoption of GPUs in the field of image processing and, more specifically, in the field of remote sensing applications aimed at the development of a favorable framework for the synergy between research and business. To fully exploit the potential of GPUs, it is essential to modify the processing algorithms by considering appropriate libraries that could be customized for the processing of geospatial data. This activity can only be carried out by combining core skills concerning the hardware and the development of libraries and algorithms specific for the wide business opportunities of the each relevant market.”
The Challenge is designed to promote integration of transversal skills useful for the development of processing algorithms specifically optimized to maximize the processing capabilities of the latest graphics boards. To this end, the Challenge’s organizers selected DORIS open source software as the target for performance enhancement. DORIS is specific for the processing of remote sensing images and is composed of a number of computing processes that lend themselves to optimization and porting to GPU platforms. Teams that join the Challenge will have the opportunity to modify these computing processes in order to reduce the overall processing execution time. The modification will be carried out in one or more steps through redesign of the algorithms in order to better exploit the GPU platform’s potential.
The output generated by the redesigned computing chain must be consistent with the reference output provided to participants and produced by using the original DORIS version. The expected result is a drastic reduction in processing execution time without any substantial change in the output.
How to apply
- To participate in the Challenge, teams may sign up through the Web site www.planetek.it/eng/gpu4eo.
- Participants should download a copy of DORIS software from http://doris.tudelft.nl/.
- On the same site, all of the information for installation and configuration of the software is available.
Both input data (SAR interferometric pair and DEM of the area of interest) and expected output data (Differential interferogram and Coherence Map) will be made available to subscribers. The developments must be optimized for NVIDIA K40 GPU. A reference hardware platform will be made available to participants by ReCaS data center; it is based on the NVIDIA K40 board and will be exploitable for testing the developments and assessing the performance achieved.
The evaluation committee
The committee is composed of representatives from the world of research and industry. Members of the 2015 evaluation committee are:
- Cristoforo Abbattista, Planetek Italia srl
- Giacinto Donvito, Datacenter ReCas-INFN
- Francesca De Leo, CNR
- Raffaele Nutricato, GAP
- Alex Pompili, Uniba
Organization of the event
The “Challenge GPU4EO” is organized by Planetek Italia srl, as part of the Project ““RIESCO – Ricerca ImprESa COnoscenza: l’innovazione dal laboratorio all’impresa in Puglia””, which involves CNR, ENEA, INFN with the participation of Confindustria Puglia. Partners in the initiative are GAP LTD to support use of the software DORIS, INFN and the Department of Physics, University of Bari who have provided the reference platform in the datacenter RECAS.
- The full Challenge announcement is available in PDF at http://www.planetek.it/eng/gpu4eo
Requests for additional information or support can be sent via e-mail to gpu4eo@planetek.it - The contest is open until June 2, 2015.
- Closing: Participants must send the activities report to gpu4eo@planetek by June 15, 2015.
- Award ceremony: June 25, 2015