Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 6
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

PreVIous: A Methodology for Prediction of Visual Inference Performance on IoT Devices

  • Velasco-Montero, Delia
  • Fernández-Berni, J.
  • Carmona-Galán, R.
  • Rodríguez-Vázquez, Ángel
This paper presents PreVIous, a methodology to predict the performance of Convolutional Neural Networks (CNNs) in terms of throughput and energy consumption on vision-enabled devices for the Internet of Things. CNNs typically constitute a massive computational load for such devices, which are characterized by scarce hardware resources to be shared among multiple concurrent tasks. Therefore, it is critical to select the optimal CNN architecture for a particular hardware platform according to prescribed application requirements. However, the zoo of CNN models is already vast and rapidly growing. To facilitate a suitable selection, we introduce a prediction framework that allows to evaluate the performance of CNNs prior to their actual implementation. The proposed methodology is based on PreVIousNet, a neural network specifically designed to build accurate per-layer performance predictive models. PreVIousNet incorporates the most usual parameters found in state-of-the-art network architectures. The resulting predictive models for inference time and energy have been tested against comprehensive characterizations of seven well-known CNN models running on two different software frameworks and two different embedded platforms. To the best of our knowledge, this is the most extensive study in the literature concerning CNN performance prediction on low-power low-cost devices. The average deviation between predictions and real measurements is remarkably low, ranging from 3% to 10%. This means state-of-the-art modeling accuracy. As an additional asset, the fine-grained a priori analysis provided by PreVIous could also be exploited by neural architecture search engines., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Contribución de congreso (ConferenceOutput)

VersaTile Convolutional Neural Network Mapping on FPGAs

  • Muñío-Gracia, A.
  • Fernández-Berni, J.
  • Carmona-Galán, R.
  • Rodríguez-Vázquez, Ángel
Convolutional Neural Networks (ConvNets) are directed acyclic graphs with node transitions determined by a set of configuration parameters. In this paper, we describe a dynamically configurable hardware architecture that enables data allocation strategy adjustment according to ConvNets layer characteristics. The proposed flexible scheduling solution allows the accelerator design to be portable across various scenarios of computation and memory resources availability. For instance, FPGA block-RAM resources can be properly balanced for optimization of data distribution and minimization of off-chip memory accesses. We explore the selection of tailored scheduling policies that translate into efficient on-chip data reuse and hence lower energy consumption. The system can autonomously adapt its behavior with no need of platform reconfiguration nor user supervision. Experimental results are presented and compared with state-of-the-art accelerators., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

A Low-Resources TDC for Multi-Channel Direct ToF Readout Based on a 28-nm FPGA

  • Parsakordasiabi, Mojtaba
  • Vornicu, Ion
  • Rodríguez-Vázquez, Ángel
  • Carmona-Galán, R.
© 2021 by the authors., In this paper, we present a proposed field programmable gate array (FPGA)-based timeto- digital converter (TDC) architecture to achieve high performance with low usage of resources. This TDC can be employed for multi-channel direct Time-of-Flight (ToF) applications. The proposed architecture consists of a synchronizing input stage, a tuned tapped delay line (TDL), a combinatory encoder of ones and zeros counters, and an online calibration stage. The experimental results of the TDC in an Artix-7 FPGA show a differential non-linearity (DNL) in the range of [0.953, 1.185] LSB, and an integral non-linearity (INL) within [2.750, 1.238] LSB. The measured LSB size and precision are 22.2 ps and 26.04 ps, respectively. Moreover, the proposed architecture requires low FPGA resources., This work was supported by EU H2020 MSCA through Project ACHIEVE-ITN (Grant No 765866), by the Spanish MINECO and European Region Development Fund (ERDF/FEDER) through Project RTI2018-097088-B-C31, and by the US Office of Naval Research through Grant No. N00014-19-1-2156., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

A customizable thermographic imaging system for medical image acquisition and processing

  • Leñero-Bardallo, J. A.
  • Rosa-Vidal, Rafael de la
  • Padial-Allué, Rubén
  • Ceballos-Cáceres, J.
  • Rodríguez-Vázquez, Ángel
  • Bernabeu-Wittel, José
A custom system has been developed for medical image acquisition and processing in both the visible and the infrared (IR) bands. Unlike some non-customizable commercial devices, this system can easily be adapted to different application scenarios by adding new peripherals and/or custom image processing algorithms. Portable, autonomous, and easy to use, it offers immediate results at the moment of examination and has a competitive cost. The system comprises a Single Board Computer (SBC) controlling a group of sensors and peripherals. The hardware implementation, described in detail in this paper, was adapted for two different application scenarios. First, the system was employed to differentiate between different kinds of vascular anomalies. The clinical results obtained are reported. The device was then redesigned to automatically detect people with high body temperatures in public environments. The system’s real-time image processing capabilities in both scenarios are demonstrated. Specific algorithms were implemented by the authors for each case study., Ayudas a Proyectos de I+D+I Programa Operativo Fondo Europeo de Desarrollo Regional (FEDER) (Grant Number: US-1264940). Proyecto Singular de Transferencia del Conocimiento: Ecosistema Innovador con Inteligencia Artifial para Andalucía 2025 (Grant Number: CEI-07). 10.13039/501100003329-Spanish Government Ministerio de Economía y Competitividad (MINECO) and European Regional Development Fund (ERDF/FEDER) (Grant Number: RTI2018-097088-B-C31). 10.13039/100007297-Office of Naval Research (ONR) Naval International Cooperative Opportunities in Science and Technology (NICOP) Program (Grant Number: N00014-19-1-2156). EU H2020 Marie Skłodowska-Curie Actions (MSCA) ACHIEVE-ITN (Grant Number: 765866), Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

An efficient TDC using a dual-mode resource-saving method evaluated in a 28-nm FPGA

  • Parsakordasiabi, Mojtaba
  • Vornicu, Ion
  • Rodríguez-Vázquez, Ángel
  • Carmona-Galán, R.
FPGA-based time-to-digital converters (TDCs) are required to be accurate, linear, and fast, while at the same time employing a reduced number of resources. Pushing these requirements to the limit is challenging, although it is constantly required by many applications. This article presents a dual-mode tapped-delay-line (TDL)—propagating 1’s and 0’s in alternating measurement cycles—architecture for a field-programmable gate array (FPGA)-based TDC that complies with the mentioned specifications. The dead-time of the proposed TDC is reduced to one system clock cycle by using a toggling input stage and a dual-mode counter-based encoder. To improve the TDC linearity, the TDL sampling sequence is tuned separately for each operating mode. The presented architecture employs a low-resources dual-mode combinatory encoder of one- and zero-counters to remove the bubbles and cover both operating modes. A dual-mode bin-width calibration has been carried out to improve the TDC performance in each mode. The proposed architecture has been implemented on a Xilinx Artix-7 FPGA. Experimental results have shown a differential nonlinearity (DNL) within [−0.71 1.05] least significant bit (LSB) and an integral nonlinearity (INL) within [−0.85 0.86] LSB for the propagation of 1’s. DNL and INL are within [−0.73 1.06] LSB and [−1.17 0.04] LSB, respectively, for the propagation of 0’s. The LSB size is 22.1 ps and the TDC precision is 22.35 ps. A comparison with recently published state-of-the-art FPGA-based TDCs is provided at the end of the article., EU H2020 MSCA through Project ACHIEVE-ITN (Grant Number: 765866). Spanish MINECO and European Region Development Fund (ERDF/FEDER) (Grant Number: RTI2018-097088-B-C31). 10.13039/100000006-U.S. Office of Naval Research (Grant Number: N00014-19-1-2156)., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Architecture-level optimization on digital silicon photomultipliers for medical imaging

  • Bandi, Franco
  • Ilisie, Víctor
  • Vornicu, Ion
  • Carmona-Galán, R.
  • Benlloch Baviera, José María
  • Rodríguez-Vázquez, Ángel
This article belongs to the Section Physical Sensors., Silicon photomultipliers (SiPMs) are arrays of single-photon avalanche diodes (SPADs) connected in parallel. Analog silicon photomultipliers are built in custom technologies optimized for detection efficiency. Digital silicon photomultipliers are built in CMOS technology. Although CMOS SPADs are less sensitive, they can incorporate additional functionality at the sensor plane, which is required in some applications for an accurate detection in terms of energy, timestamp, and spatial location. This additional circuitry comprises active quenching and recharge circuits, pulse combining and counting logic, and a time-to-digital converter. This, together with the disconnection of defective SPADs, results in a reduction of the light-sensitive area. In addition, the pile-up of pulses, in space and in time, translates into additional efficiency losses that are inherent to digital SiPMs. The design of digital SiPMs must include some sort of optimization of the pixel architecture in order to maximize sensitivity. In this paper, we identify the most relevant variables that determine the influence of SPAD yield, fill factor loss, and spatial and temporal pile-up in the photon detection efficiency. An optimum of 8% is found for different pixel sizes. The potential benefits of molecular imaging of these optimized and small-sized pixels with independent timestamping capabilities are also analyzed., This work was supported by EU H2020 MSCA through Project ACHIEVE-ITN (Grant No 765866), by the Spanish MINECO and European Region Development Fund (ERDF/FEDER) through Project RTI2018-097088-B-C31, by the US Office of Naval Research through Grant No. N00014-19-1-2156, by the European Research Council through Grant No. 695536, and by Generalitat Valenciana APOSTD/2019/086., Peer reviewed