A more complete pipeline to read text in natural scene images (in the wild). It uses multiple Deep Learning methods and sets the emphasis mostly on the preprocessing of text images to improve the performance of the OCR.
As i said on title i cant get more than one boxes for image, if image get different shapes it should work, but i mean as on example, u r just getting box of 'West', u dont get 'East' and 'Line'
I know it should like implement for on run_east_model but i dont see how.