publication . Conference object . 2016

A heterogeneous multi-core system-on-chip for energy efficient brain inspired vision

Antonio Pullini; Francesco Conti; Davide Rossi; Igor Loi; Michael Gautschi; Luca Benini;
Closed Access English
  • Published: 01 Nov 2016
  • Publisher: Institute of Electrical and Electronics Engineers Inc.
  • Country: Italy
Computer vision (CV) based on Convolutional Neural Networks (CNN) is a rapidly developing field thanks to CNN's flexibility, strong generalization capability and classification accuracy (matching and sometimes exceeding human performance). CNN-based classifiers are typically deployed on servers or high-end embedded platforms. However, their ability to “compress” low information density data such as images into highly informative classification tags makes them extremely interesting for wearable and IoT scenarios, should it be possible to fit their computational requirements within deeply embedded devices such as visual sensor nodes. We propose a 65nm system-on-chip implementing a hybrid HW/SW CNN accelerator while meeting this energy efficiency target. The SoC integrates a near-threshold parallel processor cluster [1] and a hardware accelerator for convolution-accumulation operations [2], which constitute the basic kernel of CNNs: it achieves peak performance of 11.2 GMAC/s @ 1.2 V and peak energy efficiency of 261 GMAC/s/W @ 0.65V.
Persistent Identifiers
Sustainable Development Goals (SDG) [Beta]
free text keywords: Electrical and Electronic Engineering, System on a chip, Server, Computer engineering, Efficient energy use, Multi-core processor, Kernel (image processing), Convolutional neural network, Embedded system, business.industry, business, Computer science, Hardware acceleration
Communities with gateway
OpenAIRE Connect image
Funded by
EC| ExaNoDe
European Exascale Processor Memory Node Design
  • Funder: European Commission (EC)
  • Project Code: 671578
  • Funding stream: H2020 | RIA
Validated by funder
Any information missing or wrong?Report an Issue