pytesseract: image_to_data returns different results than image_to_string

I use image_to_string for single digits but additionally I want to obtain the confidence value. Therefore I replaced it with image_to_data like this:

ocrResult = pytesseract.image_to_data(digitBinary, config='-psm 10 -c tessedit_char_whitelist=0123456789', output_type="dict")
digitasNumber = ocrResult["text"][0]

With image_to_string the results are reasonably good. With image_to_data every text dict entry is empty but for one case, where it returns two digits. the empty digits have conf -1 and the digit where text is filled the conf is 6.

I don’t think that this is intended behavior.

About this issue

Original URL
State: closed
Created 6 years ago
Comments: 19 (6 by maintainers)

Commits related to this issue

Add image path support and fix ordering for the config handling Now we can pass the image path as a string and it will be passed to tesseract directly without creating temporary image. Fixes #127 Al... — committed to madmaze/pytesseract by int3l 6 years ago

Most upvoted comments

The default value is not the real problem here. The problem is that the tsv config is apparently loaded after the config is applied (see my use case in the first post) and overwriting the value. I tried to fix the code by myself but I didn’t manage to make it work so far.

Internally it is currently called like this

"tesseract" + config + image +" tsv"

but it should be

"tesseract " + config + " -c tessedit_create_tsv=1 "+ image

BSVogler on Jul 15, 2018