A project to solve a Sudoku from a given image.

Image file with unsolved Sudoku

The same image file with filled in solution

Images currently can not be rotated so only images with perfect rotation are processed. Only images with only one Sudoku and no other artifacts can be processed. The Sudoku does not have to be fullscreen as a single Sudoku can be detected and cropped. Each detected Sudoku is scaled to 500x500 pixels and stored like this.

Provides functions to solve a sudoku puzzle.

Sudokus are formatted as a string where the rows are concatenated and empty spaces are represented as "0" or as ".".

"100200040020003900907000500004000057000541000350000100003000709001400080090002006"

Takes an image of a unresolved Sudoku puzzle as input, detects the existing digits, solves the Sudoku and returns the given image with the solution filled in. The output file is stored to the same directory as the input and has "_solved" concatenated to the image name.

python3 solve_sudoku_from_image.py --image <path>

OpenCV2 is used for the image processing.

Tesseract is used to detect the digits in the given image.