i never worked with opencv and tesseract before, so writing the backend that lines a [possibly weirdly scanned] fax up to a nice template and then performing text/checkbox detection on it was a lot of fun