A reconfigurable accelerator for neuromorphic object recognition

Jagdish Sabarad, Srinidhi Kestur, Mi Sun Park, Dharav Dantara, Vijaykrishnan Narayanan, Yang Chen, Deepak Khosla

Research output: Chapter in Book/Report/Conference proceedingConference contribution

18 Citations (Scopus)

Abstract

Advances in neuroscience have enabled researchers to develop computational models of auditory, visual and learning perceptions in the human brain. HMAX, which is a biologically inspired model of the visual cortex, has been shown to outperform standard computer vision approaches for multi-class object recognition. HMAX, while computationally demanding, can be potentially applied in various applications such as autonomous vehicle navigation, unmanned surveillance and robotics. In this paper, we present a reconfigurable hardware accelerator for the time-consuming S2 stage of the HMAX model. The accelerator leverages spatial parallelism, dedicated wide data buses with on-chip memories to provide an energy efficient solution to enable adoption into embedded systems. We present a systolic array-based architecture which includes a run-time reconfigurable convolution engine which can perform multiple variable-sized convolutions in parallel. An automation flow is described for this accelerator which can generate optimal hardware configurations for a given algorithmic specification and also perform run-time configuration and execution seamlessly. Experimental results on Virtex-6 FPGA platforms show 5X to 11X speedups and 14X to 33X higher performance-per-Watt over a CNS-based implementation on a Tesla GPU.

Original languageEnglish (US)
Title of host publicationASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference
Pages813-818
Number of pages6
DOIs
StatePublished - Apr 26 2012
Event17th Asia and South Pacific Design Automation Conference, ASP-DAC 2012 - Sydney, NSW, Australia
Duration: Jan 30 2012Feb 2 2012

Publication series

NameProceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

Other

Other17th Asia and South Pacific Design Automation Conference, ASP-DAC 2012
CountryAustralia
CitySydney, NSW
Period1/30/122/2/12

Fingerprint

Object recognition
Particle accelerators
Convolution
Reconfigurable hardware
Systolic arrays
Embedded systems
Computer vision
Field programmable gate arrays (FPGA)
Brain
Navigation
Robotics
Automation
Engines
Specifications
Hardware
Data storage equipment

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design
  • Electrical and Electronic Engineering

Cite this

Sabarad, J., Kestur, S., Park, M. S., Dantara, D., Narayanan, V., Chen, Y., & Khosla, D. (2012). A reconfigurable accelerator for neuromorphic object recognition. In ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference (pp. 813-818). [6165067] (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). https://doi.org/10.1109/ASPDAC.2012.6165067
Sabarad, Jagdish ; Kestur, Srinidhi ; Park, Mi Sun ; Dantara, Dharav ; Narayanan, Vijaykrishnan ; Chen, Yang ; Khosla, Deepak. / A reconfigurable accelerator for neuromorphic object recognition. ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference. 2012. pp. 813-818 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).
@inproceedings{655c4e52c9df43d3ae823035099ccb53,
title = "A reconfigurable accelerator for neuromorphic object recognition",
abstract = "Advances in neuroscience have enabled researchers to develop computational models of auditory, visual and learning perceptions in the human brain. HMAX, which is a biologically inspired model of the visual cortex, has been shown to outperform standard computer vision approaches for multi-class object recognition. HMAX, while computationally demanding, can be potentially applied in various applications such as autonomous vehicle navigation, unmanned surveillance and robotics. In this paper, we present a reconfigurable hardware accelerator for the time-consuming S2 stage of the HMAX model. The accelerator leverages spatial parallelism, dedicated wide data buses with on-chip memories to provide an energy efficient solution to enable adoption into embedded systems. We present a systolic array-based architecture which includes a run-time reconfigurable convolution engine which can perform multiple variable-sized convolutions in parallel. An automation flow is described for this accelerator which can generate optimal hardware configurations for a given algorithmic specification and also perform run-time configuration and execution seamlessly. Experimental results on Virtex-6 FPGA platforms show 5X to 11X speedups and 14X to 33X higher performance-per-Watt over a CNS-based implementation on a Tesla GPU.",
author = "Jagdish Sabarad and Srinidhi Kestur and Park, {Mi Sun} and Dharav Dantara and Vijaykrishnan Narayanan and Yang Chen and Deepak Khosla",
year = "2012",
month = "4",
day = "26",
doi = "10.1109/ASPDAC.2012.6165067",
language = "English (US)",
isbn = "9781467307727",
series = "Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC",
pages = "813--818",
booktitle = "ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference",

}

Sabarad, J, Kestur, S, Park, MS, Dantara, D, Narayanan, V, Chen, Y & Khosla, D 2012, A reconfigurable accelerator for neuromorphic object recognition. in ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference., 6165067, Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, pp. 813-818, 17th Asia and South Pacific Design Automation Conference, ASP-DAC 2012, Sydney, NSW, Australia, 1/30/12. https://doi.org/10.1109/ASPDAC.2012.6165067

A reconfigurable accelerator for neuromorphic object recognition. / Sabarad, Jagdish; Kestur, Srinidhi; Park, Mi Sun; Dantara, Dharav; Narayanan, Vijaykrishnan; Chen, Yang; Khosla, Deepak.

ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference. 2012. p. 813-818 6165067 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - A reconfigurable accelerator for neuromorphic object recognition

AU - Sabarad, Jagdish

AU - Kestur, Srinidhi

AU - Park, Mi Sun

AU - Dantara, Dharav

AU - Narayanan, Vijaykrishnan

AU - Chen, Yang

AU - Khosla, Deepak

PY - 2012/4/26

Y1 - 2012/4/26

N2 - Advances in neuroscience have enabled researchers to develop computational models of auditory, visual and learning perceptions in the human brain. HMAX, which is a biologically inspired model of the visual cortex, has been shown to outperform standard computer vision approaches for multi-class object recognition. HMAX, while computationally demanding, can be potentially applied in various applications such as autonomous vehicle navigation, unmanned surveillance and robotics. In this paper, we present a reconfigurable hardware accelerator for the time-consuming S2 stage of the HMAX model. The accelerator leverages spatial parallelism, dedicated wide data buses with on-chip memories to provide an energy efficient solution to enable adoption into embedded systems. We present a systolic array-based architecture which includes a run-time reconfigurable convolution engine which can perform multiple variable-sized convolutions in parallel. An automation flow is described for this accelerator which can generate optimal hardware configurations for a given algorithmic specification and also perform run-time configuration and execution seamlessly. Experimental results on Virtex-6 FPGA platforms show 5X to 11X speedups and 14X to 33X higher performance-per-Watt over a CNS-based implementation on a Tesla GPU.

AB - Advances in neuroscience have enabled researchers to develop computational models of auditory, visual and learning perceptions in the human brain. HMAX, which is a biologically inspired model of the visual cortex, has been shown to outperform standard computer vision approaches for multi-class object recognition. HMAX, while computationally demanding, can be potentially applied in various applications such as autonomous vehicle navigation, unmanned surveillance and robotics. In this paper, we present a reconfigurable hardware accelerator for the time-consuming S2 stage of the HMAX model. The accelerator leverages spatial parallelism, dedicated wide data buses with on-chip memories to provide an energy efficient solution to enable adoption into embedded systems. We present a systolic array-based architecture which includes a run-time reconfigurable convolution engine which can perform multiple variable-sized convolutions in parallel. An automation flow is described for this accelerator which can generate optimal hardware configurations for a given algorithmic specification and also perform run-time configuration and execution seamlessly. Experimental results on Virtex-6 FPGA platforms show 5X to 11X speedups and 14X to 33X higher performance-per-Watt over a CNS-based implementation on a Tesla GPU.

UR - http://www.scopus.com/inward/record.url?scp=84859945430&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84859945430&partnerID=8YFLogxK

U2 - 10.1109/ASPDAC.2012.6165067

DO - 10.1109/ASPDAC.2012.6165067

M3 - Conference contribution

AN - SCOPUS:84859945430

SN - 9781467307727

T3 - Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

SP - 813

EP - 818

BT - ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference

ER -

Sabarad J, Kestur S, Park MS, Dantara D, Narayanan V, Chen Y et al. A reconfigurable accelerator for neuromorphic object recognition. In ASP-DAC 2012 - 17th Asia and South Pacific Design Automation Conference. 2012. p. 813-818. 6165067. (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). https://doi.org/10.1109/ASPDAC.2012.6165067