I am trying to read text, mainly numbers(dates) from images that are in the form of scanned images. I am attaching a sample image for reference. With this type of image, I am getting about 90% accuracy but, I need it to be 100%.
For that first, I had tried different methods to improve image quality(deblurring, adjusting resolution, dynamic zooming, etc...).
Then I moved to train tesseract, for that, I followed this tutorial here
. But in this when creating a shapetable file, getting this error
"class_id >= 0 && class_id < unicharset_size_:Error:Assert failed:in file src/training/common/trainingsampleset.cpp, line 581".
I tried other tutorials too, but with no success.
It is currently reading 1/3 as V3, 1/7 as 1n, 2/3 as 2B and so on.