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.

[BSVogler]

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

[mbdilaver]

CLI prints the same output of image_to_string() to a .txt file and image_to_data() to a .tsv file when I gave parameter -c tessedit_create_tsv=1.

The problem is image_to_string() output is really good, but it doesn’t have text coordinates. image_to_data() output has all of the additional data but it shows each word in a seperate field.

Let’s say I have an image with Text: Hello World! text on it. image_to_string() returns Text: Hello World! (or something similar to it) where image_to_data() returns ["Text:", "Hello", "World!"] (basically) and loses spaces between words. I guess I need to concatnate these words manually 😞 This is the difference I’m talking about. The result is same in terms of letters.