Parsing a Sudoku puzzle

This is a very interesting test case. As you guessed correctly, single digit numbers are difficult to OCR. Indeed, our current engine does not detect any of them.

The good news is that our upcoming new OCR engine2 detects all of them:

To get this result, I reduced the image size by 50%. This makes the number font size more normal/standard. With the original size, it missed two numbers.

The new engine will be available for public beta either this week or next.
The new engine is available for public beta :grinning: