james-see / opencv-threshold-squares-example

Chatgpt helped write this code.

example opencv threshold filter tiles

This repo takes in a folder of png images and splits them up into grids per image and identifies light areas in the grid and creates output images with the grid overlaid as well as a csv of the stats.

quickstart

1. install opencv-python via pip3 install opencv-python
2. run process.py or process-nongreen.py for use case
3. check the output.csv and the output_folder
4. Adjust threshholds as needed for use case

defaults

Threshholds

process script for lightness filter

The threshhold is set at 180, which seems to give a decent amount of lightness identification, 200 seemed to only identify 2 50px x 50px light spots per image

process script for green filter

The threshhold is set as 0.2 * average greenness of each photo to account for the relative denseness of each and adjust, since some photos are naturally greener.

Grid size

Grid size is configurable, right now it is set at 50px, assuming an average photo size is 750px x 750px 25px seems to also be a good default.

pre-processing

The script assumes all images are png. There is a script in the pre-process folder that converts any image to png and crops to square and moves to the input folder for you.

output

The images are output to output_folder and a csv of the image name, the white square count, and the total squares is output to output.csv

ChatGPT

prompts in order (each one I executed, copied the code and then tweaked it some more):

i need opencv python code to split up a png image into 50px grid and identify the light and dark squares and fill in the squares in white that are lighter than most. the image is 750x750px

I need it to also print the total number of white squares in the bottom right on the image

i need to save the image out instead of opening it and store the image name, total white squares, and total squares in a csv called output.csv instead of showing it

i need to iterate through an input folder of png images and do the same thing for each with the images numbered output1.png output2.png etc.

i need it to not do the threshhold based on the image average but by a static lightness value

instead of the lightness threshhold, I need a not-greeness threshhold, areas that are not shades of green

I needs to figure out the average green per photo instead of using a static threshhold

and that is all!

Example output at 50px squares