Frames-to-AER efficiency study based on CPUs performance counters
Year: 2010
Type: conference-paper
Abstract: Image processing in digital computer systems \nusually considers the visual information as a sequence of frames. \nThese frames are from photographs that capture reality for a \nshort period of time. They are renewed and transmitted at a rate \nof 25-30 frames per second, in a typical real-time scenario. Each \nof these frames needs to be filtered and processed in order to \ndetect a feature on it. This processing is usually based on very \nexpensive operations in terms of resource consumption \n(processor resources and processing time) for an efficient real time application. In contrast, neuro-inspired systems, which \nwork in a manner similar to the nervous system, may resolve \nthose and others more complex problems, such as visual \nrecognition in real-time. The spike-based philosophy for visual \ninformation processing based on the neuro-inspired Address Event- Representation (AER) is achieving nowadays very high \nperformances. Address-Event-Representation (AER) is a \nneuromorphic interchip communication protocol that allows for \nreal-time virtual massive connectivity between huge numbers of \nneurons located on different chips. When building multi-chip \nmuti-layered AER systems it is absolutely necessary to have a \ncomputer interface that allows (a) to read AER interchip traffic \ninto the computer and visualize it on screen, and (b) convert \nconventional frame-based video stream in the computer into \nAER and inject it at some point of the AER structure. This is \nnecessary for test and debugging of complex AER systems. A set \nof software methods for converting digital frames into AER \nformat are present in the literature. In this work we study the \nlow level performance lacks of these methods monitoring internal \nperformance hardware counters for an Intel Core 2 Quad. We \ndiscuss the results obtained and we propose improvements for \nthose software methods that did not achieve real-time properties
Source: idUS (Universidad de Sevilla)
Authors: Manuel Domínguez-Morales, Pablo Iñigo, Juan Luis Font, D. Cascado, G. Jiménez, Fernando Díaz, José Luis Sevillano, A. Linares-Barranco
Institution: Universidad de Sevilla
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1766864849
