Project Details

Algorithms, Design Methods, and Many-Core Execution Platform for Low-Power Massive Data-Rate Video and Image Processing

Project Period: 1. 4. 2014 – 30. 6. 2017

Project Type: grant

Code: 7H14002

Agency: Ministerstvo školství, mládeže a tělovýchovy ČR

Program: Společné technologické iniciativy

Czech title
Algoritmy, metody návrhu a platforma pro many-core zpracování obrazu a videa s velkou propustností a malou spotřebou energie
Type
grant
Keywords

multi-core, many-core, low-power, massive data rate processing, embedded system, platform, architecture, system software stack, algorithm, resilience, variations, composable, cross-domain, tool chain, design method, design tool, application framework, resource management, power management, hardware-software collaborative techniques, mobile handset, healthcare, security, surveillance

Abstract

In order to address new societal challenges in the various fields (especially "Healthcare and wellbeing" and "Green, safe and supportive transportation"), platform solutions for high-end image/video processing need to be brought to the next level. The ALMARVI project aims at proactively addressing these issues at various system layers. ALMARVI aims at providing cross-domain many-core platform solution, system software stack, tool chain, and adaptive algorithms that will enable massive data-rate image/video processing with high energy efficiency. ALMARVI will provide mechanisms and support for high degree of adaptivity at various system layers that will abstract the variations in the underlying platforms (e.g., due to imperfections in the fabrication process), communication channels (e.g., available bandwidth), application behaviour (dynamic workloads, changing requirements) from the application developer. This is crucial for providing consistent performance efficiency in an interoperable manner when considering heterogeneous platform options and dynamic operating conditions. The key is to leverage image/video content-specific properties, application-specific features, and inherent resilience properties of image/video processing applications. The goal of ALMARVI is to develop:    - Adaptive, scalable, and paralleled algorithms for image and video processing    - Cross-domain system software stack with adaptive run-time system for      efficient resource/power management and improved interoperability    - Concepts for continuous hardware and software adaptations    - Cross-domain many-core execution platform scalable with off-the-shelf      heterogeneous acceleration fabrics like FPGAs, embedded GPUs, DSPs, etc.    - Design tools and methods for execution platform    - Industrial-grade demonstrators for multiple application use cases to      validate the project results.      ALMARVI's scientific and technological developments will be validated and      demonstrated in diverse use cases from three key industrial domains i.e.,      healthcare, security/surveillance/monitoring, and mobile. Therefore, in addition to      providing scalable and cross-domain computing platforms for embedded systems, ALMARVI aims      at providing "the core of solutions for the big societal challenges like affordable healthcare      and wellbeing, green and safe transportation, reduced consumption of power ... [Artemis JU Annual      Work Programme (AWP) 2013, Section 1] [28]". ALMARVI's technological developments will help in combating      the complexity of advanced image/video processing applications from key industrial domains of European market,      thus providing improved productivity and increased competitiveness of European market in different sectors.      ALMARVI's platform solution will enable development of low-cost solutions for a wide-range of market in different      industrial domains and create new market opportunities, in particular supporting SMEs. The ALMARVI project exhibits      a consortium with complementary skills, market/industrial/business experience, and technical/organizational competence. This is      crucial to accomplish the challenging goals of ALMARVI. ALMARVI partners are highly active at EU and international levels/markets. They      possess a vast experience in contributing, successfully executing, and managing EU R&D projects.

Team members
Zemčík Pavel, prof. Dr. Ing., dr. h. c. (DCGM) – research leader
Čekan Ondřej, Ing., Ph.D. (UFYZ)
Hradiš Michal, Ing., Ph.D. (DCGM)
Hruška Tomáš, prof. Ing., CSc. (DIFS)
Jašek Roman, Ing. (FIT)
Juráček Ivo, Bc.
Kolář Dušan, doc. Dr. Ing. (DIFS)
Kotásek Zdeněk, doc. Ing., CSc.
Krčma Martin, Ing., Ph.D. (UFYZ)
Lojda Jakub, Ing., Ph.D. (DCSY)
Podivínský Jakub, Ing., Ph.D. (UFYZ)
Smrž Pavel, doc. RNDr., Ph.D. (DCGM)
Svoboda Pavel, Ing., Ph.D.
Szentandrási István, Ing., Ph.D.
Zachariáš Michal, Ing., Ph.D. (DCGM)
Zachariášová Marcela, Ing., Ph.D. (DCSY)
Publications

2019

  • VEĽAS, M.; ŠPANĚL, M.; HRADIŠ, M.; HEROUT, A. Convolutional Neural Networks for the Odometry Estimation. Journal of Intelligent and Robotics Systems, 2019, vol. 2019, no. 93, p. 1-22. ISSN: 0921-0296. Detail

2017

  • ČEKAN, O.; KOTÁSEK, Z. Random Test Stimuli Generation Based on a Probabilistic Grammar. Proceedings of the 5th Prague Embedded Systems Workshop. Roztoky u Prahy: Faculty of Information Technology, Czech Technical University, 2017. p. 43-44. ISBN: 978-80-01-06178-7. Detail
  • ČEKAN, O.; KOTÁSEK, Z. A Probabilistic Context-Free Grammar Based Random Test Program Generation. In Proceedings of 20th Euromicro Conference on Digital System Design. Vídeň: Technical University Wien, 2017. p. 356-359. ISBN: 978-1-5386-2145-5. Detail
  • KRČMA, M.; KOTÁSEK, Z. Approximation accuracy of different FPNN types. Proceedings of the 5th Prague Embedded Systems Workshop. Roztoky u Prahy: Faculty of Information Technology, Czech Technical University, 2017. p. 81-82. ISBN: 978-80-01-06178-7. Detail
  • LOJDA, J.; KOTÁSEK, Z. A Basic Approach to Fault Tolerance of Data Paths of HLS-synthesized Systems and its Evaluation. Proceedings of the 5th Prague Embedded Systems Workshop. Roztoky u Prahy: Faculty of Information Technology, Czech Technical University, 2017. p. 79-80. ISBN: 978-80-01-06178-7. Detail
  • PODIVÍNSKÝ, J.; ČEKAN, O.; LOJDA, J.; ZACHARIÁŠOVÁ, M.; KRČMA, M.; KOTÁSEK, Z. Functional Verification Based Platform for Evaluating Fault Tolerance Properties. Microprocessors and Microsystems, 2017, vol. 52, no. 5, p. 145-159. ISSN: 0141-9331. Detail
  • PODIVÍNSKÝ, J.; KOTÁSEK, Z. The Use of Functional Verification for Monitoring Impact of Faults in SRAM-based FPGAs. Proceedings of the 5th Prague Embedded Systems Workshop. Roztoky u Prahy: Faculty of Information Technology, Czech Technical University, 2017. p. 81-82. ISBN: 978-80-01-06178-7. Detail

2016

  • ČEKAN, O. Generování testovacích stimulů. Počítačové architektury a diagnostika PAD 2016. Bořetice - Kraví Hora: Fakulta informačních technologií VUT v Brně, 2016. s. 97-100. ISBN: 978-80-214-5376-0. Detail
  • ČEKAN, O.; KOTÁSEK, Z. Software-implemented Fault-Tolerant Program Generation. Proceedings of the 4th Prague Embedded Systems Workshop. Roztoky u Prahy: 2016. p. 13-13. ISBN: 978-80-01-05984-5. Detail
  • ČEKAN, O.; PODIVÍNSKÝ, J.; KOTÁSEK, Z. Random Stimuli Generation Based on a Stochastic Context-Free Grammar. In Proceedings of the 2016 International Conference on Field Programmable Technology. Xi'an: IEEE Computer Society, 2016. p. 295-296. ISBN: 978-1-5090-5602-6. Detail
  • KOLÁŘ, M.; HRADIŠ, M.; ZEMČÍK, P. Deep Learning on Small Datasets using Online Image Search. In Proceedings of 32nd Spring Conference on Computer Graphics. Proceeding of Spring Conference on Computer Graphics. Bratislava: Comenius University in Bratislava, 2016. p. 87-93. ISBN: 978-1-4503-3693-2. ISSN: 1335-5694. Detail
  • KOTÁSEK, Z.; PODIVÍNSKÝ, J. Verification of Robot Controller for Evaluating Impacts of Faults in Electro-mechanical Systems. Proceedings of the 4th Prague Embedded Systems Workshop. Roztoky u Prahy: 2016. p. 0-0. Detail
  • KRČMA, M.; KOTÁSEK, Z.; LOJDA, J.; KAŠTIL, J. Comparsion of FPNNs models approximation capabilities and resources utilization. Proceedings of the Work in progress Session held in connection with DSD 2016. Limassol: Johannes Kepler University Linz, 2016. p. 1-2. ISBN: 978-3-902457-46-2. Detail
  • LOJDA, J.; PODIVÍNSKÝ, J.; KRČMA, M.; KOTÁSEK, Z. HLS-based Fault Tolerance Approach for SRAM-based FPGAs. In Proceedings of the 2016 International Conference on Field Programmable Technology. Xi'an: IEEE Computer Society, 2016. p. 301-302. ISBN: 978-1-5090-5602-6. Detail
  • PODIVÍNSKÝ, J. Funkční verifikace jako nástroj pro sledování vlivu poruch na elektro-mechanický systém. Počítačové architektury a diagnostika PAD 2016. Bořetice - Kraví Hora: Fakulta informačních technologií VUT v Brně, 2016. s. 101-104. ISBN: 978-80-214-5376-0. Detail
  • PODIVÍNSKÝ, J.; ČEKAN, O.; LOJDA, J.; KOTÁSEK, Z. Functional Verification as a Tool for Monitoring Impact of Faults in SRAM-based FPGAs. In Proceedings of the 2016 International Conference on Field Programmable Technology. Xi'an: IEEE Computer Society, 2016. p. 293-294. ISBN: 978-1-5090-5602-6. Detail
  • PODIVÍNSKÝ, J.; ČEKAN, O.; LOJDA, J.; KOTÁSEK, Z. Verification of Robot Controller for Evaluating Impacts of Faults in Electro-mechanical Systems. In Proceedings of the 19th Euromicro Conference on Digital Systems Design. Limassol: IEEE Computer Society, 2016. p. 487-494. ISBN: 978-1-5090-2816-0. Detail
  • SVOBODA, P.; HRADIŠ, M.; BAŘINA, D.; ZEMČÍK, P. Compression Artifacts Removal Using Convolutional Neural Networks. Journal of WSCG, 2016, vol. 24, no. 2, p. 63-72. ISSN: 1213-6972. Detail
  • SVOBODA, P.; HRADIŠ, M.; MARŠÍK, L.; ZEMČÍK, P. CNN for license plate motion deblurring. In IEEE International Conference on Image Processing (ICIP). Phoenix: IEEE Signal Processing Society, 2016. p. 3832-3836. ISBN: 978-1-4673-9961-6. Detail
  • SZURMAN, K.; MIČULKA, L.; KOTÁSEK, Z. Towards a State Synchronization Methodology for Recovery Process after Partial Reconfiguration of Fault Tolerant Systems. Proceedings of the 4th Prague Embedded Systems Workshop. Roztoky u Prahy: 2016. p. 0-0. ISBN: 978-80-01-05984-5. Detail

2015

  • HRADIŠ, M.; KOTERA, J.; ZEMČÍK, P.; ŠROUBEK, F. Convolutional Neural Networks for Direct Text Deblurring. Proceedings of BMVC 2015. Swansea: The British Machine Vision Association and Society for Pattern Recognition, 2015. p. 1-13. ISBN: 1-901725-53-7. Detail
  • SZENTANDRÁSI, I.; ZACHARIÁŠ, M.; KAJAN, R.; TINKA, J.; JURÁNKOVÁ, M.; SOCHOR, J.; HEROUT, A. INCAST: Interactive Camera Streams for Surveillance Cams AR. In Proceedings of the 2015 14th IEEE International Symposium on Mixed and Augmented Reality. Fukuoka: Institute of Electrical and Electronics Engineers, 2015. p. 1-5. ISBN: 978-1-4799-6184-9. Detail
Back to top