Tesseract 5.x for Math recognition

[kwmz...@gmail.com]

Hi everyone, 

I'm trying to use Tesseract to detect both the english part and the mathematical part of the image below and it doesn't seem to work 

The code I'm using is :

import pytesseract

import cv2

custom_config = r'-l eng+equ'

img = cv2.imread("tessa.png")
text = pytesseract.image_to_string(img, config=custom_config,)
print(text)

The output being produced is just (see below) without the mathematical part even though I've used eng+equ 

Did anyone find a workaround for this or must I retrain tesseract? 

Regards,

Nash

[Đỗ Đức Phượng]

The code you provided uses Tesseract OCR with a custom configuration (-l eng+equ) to recognize English and mathematical equations (equ) in the image. However, there is a small issue with the code – pytesseract.image_to_string() expects the image in PIL (Python Imaging Library) format, not OpenCV format (NumPy array).

To fix the issue, you can convert the image from OpenCV format to PIL format before passing it to Tesseract. You can use the PIL.Image.fromarray() function to perform this conversion.

Here's the updated code:

pythonCopy code

import pytesseract import cv2 from PIL import Image custom_config = r'-l eng+equ' img = cv2.imread("tessa.png") # Convert the image from OpenCV format (NumPy array) to PIL format pil_image = Image.fromarray(img) # Perform OCR using Tesseract and extract text from the image text = pytesseract.image_to_string(pil_image, config=custom_config) print(text)

Make sure to replace "tessa.png" with the actual path to your image file.

With this code, Tesseract OCR will attempt to recognize both English text and mathematical equations present in the image. The custom_config parameter with the value -l eng+equ instructs Tesseract to use the English and mathematical equation language data for recognition.

Please note that while Tesseract is a powerful OCR engine, recognizing complex mathematical expressions accurately might be challenging. If you encounter issues with accuracy, consider using specialized OCR libraries or APIs that are designed specifically for math recognition.

source: Chat gpt

Vào lúc 04:55:05 UTC+7 ngày Thứ Tư, 19 tháng 7, 2023, kwmz...@gmail.com đã viết:

[Nash Kwmz]

I tried this method as well. I even tried performing some pre-processing to the image to give tesseract a better idea of what's going on but its still not working.