Tesseract unable to read math expression

Question

I got this image of a simple math expression Tesseract fails to read:

I've tested a screenshot of the same expression written on an Android phone and it was read pretty well. So I thought it's a font problem. I considered:

• Preprocess the image by inverting or removing the red areas
• Training Tesseract with images (StackOverflow question with no answers)
• Using WhatFontIs.com to find similar font then training Tesseract with the font file with TrainYourTesseract

Answer 1

But as I was typing the question, I looked around for more.

And this answer prompted me to double check my sanity with this VietOCR software which outputs 8-3, close enough!

Then I messed around the software and found that I could pass --psm 7 (Page Segmentation Mode 7: Treat the image as a single text line) to my script, which works well for my math expressions:

pytesseract.image_to_string(img, config='--psm 7')

List of PSMs