What is the best way to find and recognize the text of a known pattern in an image?

Question

Machine learning

Numbers are printed on a white background: always N characters from a limited alphabet that excludes confusion: no ones and small "L", no zeros and letters "O". Font - looks like a monospaced, low resolution dot matrix printer, almost Braille, single dots.

code example

I would like to find in the photo and recognize such codes. Photos from different angles, at different scales, somewhere the code occupies the entire width of the frame, somewhere it is modestly shallow obliquely in the corner of the picture. Extra shadows and highlights.

What model-network can search for a pattern in an image at different scales and at different angles, so that later in the defined quadrangle it can recognize exactly N characters?

I read about multi-scale CNN - in fact, several convolutional networks work in parallel, with a picture in different resolutions: 1:1, 1:2, 1:4. Not sure if this is the most efficient solution when the proportions of the print you are looking for are known.

In some ways, the task is similar to license plate recognition, and even easier in that the pattern, length, and background are known exactly without unnecessary details.