OCR and pytesseract detecting numbers in an image

Question

currentbid.png:

I am trying to detect the number in this image and it gives me letters or the wrong number.

This is my image i am trying to detect the number ive tried tons of stuff with greyscale and inversion using tesseract but nothing seems to work it keeps giving me letters like ADA or the wrong number like if the image said 98.7M it would give me 19 9947 )M and i think that the period is messing it up but im unable to remove it or change the font. How can I fix it or train it?

Here is my current code:

pyautogui.screenshot("bidpossible.png", region=(900, 310, 450, 60)) #bidpossible
originalImage = cv2.imread('bidpossible.png')


grayImage = cv2.cvtColor(originalImage, cv2.COLOR_BGR2GRAY)

(_, blackAndWhiteImage) = cv2.threshold(grayImage, 127, 255, cv2.THRESH_BINARY_INV)

custom_config = r'--psm 8'


text = pytesseract.image_to_string(blackAndWhiteImage, config=custom_config)
print('Extracted Text: ', text)

Johnny Cheesecutter · Accepted Answer · 2024-06-12 20:44:23Z

1

how about using filter that will try to get only light-blue color that is inside text boundary?


grayImage[(grayImage<210)] = 255
grayImage[(grayImage>210) & (grayImage<230)] = 0


# test different models, imho models 6 or 7 work better
custom_config = f'--psm 7' 
text = pytesseract.image_to_string(grayImage, config=custom_config)
print(f'Extracted Text: ', text) # 4.34m_

edited Jun 12, 2024 at 20:44

answered Jun 12, 2024 at 20:36

Johnny Cheesecutter

2,9611 gold badge16 silver badges25 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

philM Over a year ago

Thank you but now im also having trouble detecting white text with a black outline that is on a white background it spits out something close to the word. So i tried a word list to detect what word is closest to it. Which works sometimes but it is unreliable. How could i go about reading it better?

Johnny Cheesecutter Over a year ago

Could you pls add a new picture to the question? Let me know if it is more complex than the first one.

Johnny Cheesecutter Over a year ago

Honestly tessarect is not the best option for complex text recognition tasks. It could make a bit more sense to switch to Neural Nets, for example to mocr library: github.com/open-mmlab/mmocr/tree/main?tab=readme-ov-file

Collectives™ on Stack Overflow

OCR and pytesseract detecting numbers in an image

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related