@kaia the number is not obscured, so the simplest OCR should be fine. the cheese is perfectly visible, and the maze's edges have a fair bit of contrast to them, so a rather simple color correction can turn this image into something usable by a maze solver. like, i did this in three seconds. you can use the diffs between blue and green to see where the cheese is (since cheese is yellow, and yellow is NOT blue). the rat can be picked up by something like uhh tensorflow image detection?