Simple text bad results

[Tyler]

Hi,

I'm having issues getting good accuracy for some text. I have read through https://github.com/tesseract-ocr/tessdoc/blob/main/ImproveQuality.md as well as tried many different image preprocessing such as resizing, dilation/erosion, thresholding, and applying blurs. I have also used appropriate whitelisted characters but still cannot get great accuracy.

See attached some example images I am having trouble with. I am using Tesseract 4.1.1, and pytesseract 0.3.10.

I'd appreciate any pointers or ideas to get better accuracy. At this point, I am thinking I may have to fine tune the model.