I am using python's CV2 library to manipulate the image and then Tesseract for OCR but whatever I do it cannot pick up the word 'Login' word in the blue button. I have a script that tries thousands of permutations of the image (grey scaling, thresholding, negative) and then feed it to Tesseract (with various psm modes, oem modes) but none find the word 'Login'. In many permutations I see all the other text (like 'Email', 'Password', 'Forgot your password' etc') but none whatsoever find 'Login'. Generally everything is working for me when I look for other texts. Not sure if it something I am doing wrong or this is some limitation. I tried some online OCRs and they were unable to find it. Note: it isn't picking up strings that are similar to the word 'Login', it just skips it entirely. My request: full CV2 and Tesseract calls with their arguments that find the word 'Login'. Emphasizing that the specific arguments in CV2 and Tesseract calls is what I need. Continue reading...
