Project Details
Projekt
Hybrid Estimation of Non-Functional Properties (Time, Power, Energy) for Software on Embedded Systems for Image and Video Processing
Subject Area
Computer Architecture, Embedded and Massively Parallel Systems
Communication Technology and Networks, High-Frequency Technology and Photonic Systems, Signal Processing and Machine Learning for Information Technology
Communication Technology and Networks, High-Frequency Technology and Photonic Systems, Signal Processing and Machine Learning for Information Technology
Term
since 2023
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 495332681
During the past years, the complexity of image and video processing hardware, algorithms, as well as the amount of image data has increased dramatically. At the same time, the need for non-functional resources like processing time, processing power, or processing energy has increased. During the development of new algorithms, however, it is required to keep track of these resources to allow early considerations on optimizations for real-time capability and energy efficiency. To this end, a well-established method is the use of processor simulation or emulation, which, unfortunately, is usually too complex for practical applications. Simplified approaches, however, which allow faster results, usually do not provide accurate estimates. Other approaches are using application specific features for the estimation of non-functional properties and therefore abstract from the hardware point of view, while unfortunately losing general applicability. Consequently, the goal of this project is to develop a hybrid approach for the estimation of non-functional properties like time, energy, and power of image and video processing algorithms. The hybrid approach shall allow obtaining highly accurate estimates in short processing times for arbitrary algorithms. This goal shall be achieved by the combination of classic simulation or emulation methods with a fast feature-based estimation approach that has successfully been applied on specific video processing tasks.
DFG Programme
Research Grants
