Energy-Efficient High-Performance Computer Vision Systems

Abstract

Big data is a national security concern: data is a national asset. This proposed research will study big data with respect to video and images using deep convolutional neural networks (DCNNs). The current issue with deep convolutional neural networks (DCNNs) is that there is high computational complexity which requires resources, and thus these DCNNs have poor efficiency. The research proposed will tackle the poor efficiency aspect by studying cross-layer design which will focus on multiple layers of abstraction (algorithm, architecture, circuit and system) including energy-efficient data flows on spatial architectures, as energy for data movement is significantly greater than computation, energy-efficient circuits that exploit data statistics, energy-efficient algorithms that efficiently map to hardware, and energy models to enable cross-layer design of algorithms and hardware, and energy-efficient systems that minimize the cost of flexibility and scalability. The flexibility cost will be compared with the programmability cost of GPUs. The research, if successful, will improve the warfighters ability to decipher big data and maintain the Air Force s battle-space awareness.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 15, 2016
Source ID
FA95501610228

Entities

People

  • Vivienne Sze

Organizations

  • Air Force Office of Scientific Research
  • Harvard University
  • United States Air Force

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks
  • Space