I am very new to OCR and I would like to know your opinion. I have a task where I need to fetch informations from identity card (just one country) using camera. I tried to OCR some bitmaps which contain front or back of id with Tesserac for Android, but results were very bad (both performance and accuracy)
Second attempt was to use MLKit from Firebase. Here, results are very very good (in contrast to Tesserac), especially when it comes to performance, but accuracy is still not sufficient. I am corious if it is even possible to make this task done without having trained my own model for that.
What is your thoughts and experience regarding that topic? Can you spot some directions I can go to accomplish this task?
Thanks in advance
to have a very good result you need to perform some image processing before applying the OCR , you need to check the official tesseract github link for more preprocessing levels check this
for training model , you almost need it for almost not clear or handwritten text