cs 19(16): e6

Research Article

Hardware Acceleration of Computer Vision and Deep Learning Algorithms on the Edge using OpenCL

Download29 downloads
  • @ARTICLE{10.4108/eai.5-11-2019.162597,
        author={B. Mishra and D. Chakraborty and S. Makkadayil and S. D.  Patil and B. Nallani},
        title={Hardware Acceleration of Computer Vision and Deep Learning Algorithms on the Edge using OpenCL},
        journal={EAI Endorsed Transactions on Cloud Systems},
        volume={5},
        number={16},
        publisher={EAI},
        journal_a={CS},
        year={2019},
        month={11},
        keywords={CNN, OpenCL, Computer Vision, Machine Learning, Industrial Automation, FPGA, OCR, Hardware Acceleration},
        doi={10.4108/eai.5-11-2019.162597}
    }
    
  • B. Mishra
    D. Chakraborty
    S. Makkadayil
    S. D. Patil
    B. Nallani
    Year: 2019
    Hardware Acceleration of Computer Vision and Deep Learning Algorithms on the Edge using OpenCL
    CS
    EAI
    DOI: 10.4108/eai.5-11-2019.162597
B. Mishra1,*, D. Chakraborty1, S. Makkadayil1, S. D. Patil2, B. Nallani3
  • 1: Intel Corporation, Bangalore, India
  • 2: Intel Corporation, Bangalore, India during the time of writing the paper
  • 3: Worked on the project at Intel Corporation, Bangalore, India
*Contact email: bakshree.mishra@intel.com

Abstract

Machine vision using CNN is a key application in Industrial automation environment, enabling real time as well as offline analytics. A lot of processing is required in real time, and in high speed environment variable latency of data transfer makes a cloud solution unreliable. There is a need for application specific hardware acceleration to process CNNs and traditional computer vision algorithms. Cost and time-to-market are critical factors in the fast moving Industrial automation segment which makes RTL based custom hardware accelerators infeasible. This work proposes a low-cost, scalable, compute-at-the-edge solution using FPGA and OpenCL. The paper proposes a methodology that can be used to accelerate traditional as well as machine learning based computer vision algorithms.