I created image data set including 62992 images with 128x128px resolution that contains characters, numbers and symbols with four kinds of font styles. how do I train this dataset and create pre-trained model for my OCR, can you please help me?
Advertisement
Answer
If you want to train a custom OCR, I would advise studying the Tensorflow Attention OCR implementation which can be found here.
I’ve used this implementation in various projects and it gave satisfactory results in converting image to text.