OCR(Optical Character Recognition) consists of text localization + text recognition.
(text localization finds where the characters are, and text recognition reads the letters.)
CRNN works well for license plate recognition as follows.
How to Training
First, you need a lot of cropped license plate images. And in my case I expressed the number of the license plate with the image file name. (The license plate number 1234 is indicated as "1234.jpg"). (You can also define labeling with txt or csv files if you want. [(ex)0001.jpg "1234" n 0002.jpg "0000" ...)
Since I used Korean license plates, I expressed the Korean language on the license plate in English.
(example) A18sk6897 A : 서울 sk : 나
After creating training data in this way, put it in 'DB/train' directory and run training.py.