I'm looking at possibly using tesseract 4 for license plate recognition. Currently I've use downloaded the trained English database - from
https://github.com/tesseract-ocr/tessdata/ - and limited returns to numbers and capital letters only. It still isn't very good, but that's probably because there is a lot of handwritten stuff in there that I don't need. I'm wondering if someone already has a trained set for license plate characters? I've looked at the training documentation but it looks horrendous, but I possibly could use something others have written to streamline this a little (I found some links earlier). But to train tesseract (v4 in particular) apparently I will need a lot of training data as well, which I don't particularly have much of, so a traineddata file with only A-Z & 0-9 would be helpful.