Multiprocessor CA Updates 

Objective

-Algorithms Ran on the FPGA

  • Gaussian Blur

  • Sobel Filter    

-Types of Processing

-Single or Serial Processing

-Multi-processing 

Images used

  • Big  Image: 
  • Resolution: 803 X 1200
  • Small Image
  • Resolution: 192 x 192

Plan 1- Approach planned to take

  • Use PYNQ- Link- 

    PYNQ is an open-source project from Xilinx® that makes it easy to design embedded systems with Xilinx Zynq® Systems on Chips (SoCs).

    Using the Python language and libraries, designers can exploit the benefits of programmable logic and microprocessors in Zynq to build more capable and exciting embedded systems.

  • Started with the edge detection algorithm: Materials found
    • Paper-1- Link
    • paper-2 -link-canny algorithm
    • Blogs- Link- Has used HLS
    • paper3- Link
    • blog- Link-Much well written
  • Most papers have preferred using solbel filters
  • Can be used via scipy library in python

Architecture

Board Used-Ultra 96 FPGA

  • List of Features The Ultra96 Developer Kit supports the following features:
  1. Zynq UltraScale+ MPSoC ZU3EG SBVA484
  2. Storage o Micron 2 GB (512M x32) LPDDR4 Memory

PYNQ-Overlays

  • Overlays, or hardware libraries, are programmable/configurable FPGA designs that extend the user application from the Processing System of the Zynq into the Programmable Logic
  • PYNQ provides a Python interface to allow overlays in the PL to be controlled from Python running in the PS. FPGA design is a specialized task that requires hardware engineering knowledge and expertise. PYNQ overlays are created by hardware designers and wrapped with this PYNQ Python API.
  • Overlay used in the project: OpenCV overlay- link

Results Obtained- Big Image

Title Text

Remaining Work

  • Try for an even larger image
  • Find out if it's possible to tinker with an overlay
  •  

Breaking down the problem

  • Instead of using large NNs, try different types of CNN filters instead
    • Edge Detection-Kernel-solbel
    • Blurring-Kernels-Gaussian
  • For Gaussian materials found
Made with Slides.com