Hello, I would like to recognise letters with my Openmv h7. To do this, I want to divide the letter into individual fields and then compare the average grey value for the individual fields and then infer the letter.
Is it possible to do this with statistics.mean()? If so, where do I write the x and y values in, as well as the height and width of the fields, in order to get the exact mean value?
Just train a CNN using edge impulse for this purpose. Or use a CNN that does this. We have TensorFlow support.
If you dont want to build ocr project but only a presence application you can use the pixels and area thresholds of a blob tool to verify that there is something written and not ocr.
If you want to build ocr application then training is the only way but very slow. If you are allright with 1-2 frames per second its ok…