switching to easyocr - Githubissues

TKFRvisionOfficial / bszet_substitution_plan

Parsing and Image creation service of the BSZET substitution plan bot

GNU Affero General Public License v3.0

2 stars 0 forks source link

Closed PBahner closed 2 years ago

PBahner commented 2 years ago

The main improvements are:

I replaced pytesseract with easyocr which works better now.
I optimized the preprocessing of images before they are converted to text
I added a function that handles parsing mistakes.
I improved the recognition of the table (what belongs to the table and what doesn't)

All in all the text-recognition works a lot better now. It's also easier to implement adjustments for parsing mistakes.

PBahner commented 2 years ago

LGTM but i feel like this will crash production

@TKFRvisionOfficial Nooo!!!! It will work better 😂