qusetion about these two parameters' difference: text_threshold and low_text

[JeremyGe07]

I have a question regarding the parameters of the readtext method for Text Detection. There are two parameters, text_threshold and low_text, described as follows:

text_threshold (float, default = 0.7): Text confidence threshold.
low_text (float, default = 0.4): Text low-bound score.

My intuition is that the value of text_threshold should be higher than the value of low_text. Since the confidence scores of the text regions that to be saved are already higher than text_threshold, they should certainly be higher than low_text as well. I'm not quite clear on the distinction between these two parameters. Thanks for your answer.

[poojatambe]

Hi, same issue facing.
There is no clear idea about low_text (Text low-bound score) indicates. Is it text detection minimum confidence value?

By changing values of low_text and text_threshold the OCR output changes. How to choose correct values to get correct OCR detection on different resolution images and small text size?

[jepperaskdk]

I don't know what it represents, but it is being used as a threshold here:
https://github.com/JaidedAI/EasyOCR/blob/master/easyocr/craft_utils.py#L27