Sobel edge detection algorithms implemented in Python for grayscale images.

This project intends to serve as a model for future Stochastic Computing applications.

• Deterministic
  • Software
  • Hardware
• Stochastic
  • Software
    • LFSR
    • Module
    • Wrapper
  • Hardware

Remarks:

• Deterministic implementation:
  • Works as OpenCV's Sobel example without image blurring
• Stochastic implementation:
  • Hardware implementation needs to be reviewed
• Image database for testing from "CURE-OR: Challenging Unreal and Real Environment for Object Recognition", IEEE Dataport, 2019. [Online]. Available: http://dx.doi.org/10.21227/h4fr-h268. Accessed: Nov. 11, 2019.

Python dependencies:

• Python >= 3.6.7
• Libraries:
  • opencv-python
  • matplotlib
  • numpy
  • scipy
  • bitarray
  • ray (recommended)
    • for parallel processing in simulation
    • currently, only supported on Linux and MacOS
  • wheel (recommended for bitarray support)
  • setproctitle (optional)
  • psutil (optional)
  • aiohttp (optional)
  • grpcio (optional)

Hardware simulation dependencies:

• Icarus Verilog 10.1
• make (for ease of executing multiple commands)

This project currently uses stochastic circuits derived from ones synthesized with scsynth/STRAUSS

Installation recommendation:

• Newer versions of Python 3 (like 3.8.x) come with pip preinstalled. PyPi/pip is a simple package manager for Python (normally aliased as pip3)
• For this project:

pip3 install --user numpy scipy matplotlib wheel opencv-python ray bitarray setproctitle psutil aiohttp grpcio