Columns of text are being merged in OCR result. How to read text from columns without merging bounding boxes?

Hi,

I want to perform OCR on the following image:

As can be seen, there are 3 columns of the main text. Which should be 3 separate bounding boxes. But the output is merging all 3 and reading them as one, for example the output from reader.readtext(img, detail=0, paragraph=True, decoder='beamsearch') is:

Peter; Chris and Sarah were London (Tfl) to extend the extension is Thousands of delighted to join Harriet Bakerloo line from Elephant to people travel from; to and

I even tried changing the margin parameters to prevent combining the bounding boxes:

text = reader.readtext(img, detail=0, paragraph=True, decoder='beamsearch', add_margin=0, width_ths=0)

and

text = reader.readtext(img, detail=0, paragraph=True, decoder='beamsearch', add_margin=0, width_ths=1)

both produce the same output, that merge the columns.

What can I do to prevent this merging and consider these as separate bounding boxes?

Thank you

[rkcosmos]

Last code update allow you to use both x_ths and width_ths to control merging behaviour when setting paragraph=True. Setting both to zero should give you desirable outcome. You can install the latest version by pip install git+git://github.com/jaidedai/easyocr.git. Pip version will be updated soon after some tests.

[rkcosmos]

I just updated API documentation.