artificial intelligence, image recognition, ocr, image processing developement

The dataset is of different ID cards from different provinces which contain Status card and Drivers Licence Card. with this dataset we have to recognize the ID cards and extract the information from there like (first name, last name, address, gender, ID number). There is one existing model which was giving 70 percent accuracy using microsoft azure ocr. We need to check how to increase accuracy if possible. Or otherwise we can build an artificial neural network model for this if you suggest. Or if you want to use existing microsoft azure model or whatever model , i mean il go for whatever you suggest.

Another work will be we have to parse information from text (NLP)