Purpose

• It was fun to mess around with :3
• Making this public so cqql can also mess around with it

Dependencies

• opencv-python
  • all of the image stuff
• matplotlib
  • color lookup table
• (numpy)
  • used internally by opencv, but part of stdlib

Running

• python -m venv localvenv
• ./localvenv/bin/pip install opencv-python
• ./localvenv/bin/pip install matplotlib
• ./localvenv/bin/python match.py

Modifications

• If you have more/less images, change MAX global
• If you want to test rotational stuff, uncomment the line in main
• It is assumed that the shape is a darker shape on a ligher background
  • The image gets converted into grayscale and assumes
  • the shape has an intensity below 200
  • the background has an intensity above 200
• Depending on how dissimilar the shapes are, NormUpperBound should be increased

Attributions

• The test images were generated using
  • https://moonlit.technology/cqql/frost_patterns

Build Blog

• pandoc --standalone blog.md -o shape_matching.html --metadata title="Shape Matching"