- https://pillow.readthedocs.io/en/stable/
- https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_thresholding/py_thresholding.html
- https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_houghlines/py_houghlines.html
- https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_canny/py_canny.html
- https://nanonets.com/blog/ocr-with-tesseract/
These two routines are used for generating the training data used to recignize numbers extracted from Sudoku puzzle matrices.
train_extract_images.py
processes images in an input directory and extracts digits from cells as individual trimmed images and puts these images files in the output file organized in directories by the digit
train_compute_image_metrics.py
uses these individual digit images to compute the training data, which goes in the output folder.
Note, these routines need to be re-written to make better use of cv2 and numpy.