You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Problem is That Data Trained By tesseract didnt work on either regular fonts used by languages like its urdu language most comon urdu font is like
while trained data is so much different
So As i see the process is like
a picture > box data and ground data > other stuff > trained model
What i think process should be like
lvl 1
--- Input ---
language
language text ( bunch of text )
language fonts
lvl 2
--- Training Data Generation ---
Automated process
divide text small pieces
uses fonts
and simulate text pieces using font files in preview window
different color combination can be applied for precision training ( may be not necessary )
capture these preview samples automatedly as images and text data can be saved as ground truth also be usd for box files
These Steps can be automated
lvl 3
--- Final Output ---
As Data is generated by lvl 2 now we just do Training
And Final trained model be ready
We can involve a lvl for testing model for new text
Tried Youtube Stuff its freaking too much irritating for such like me inexperince person too much errors
User inputs using lvl 1 and lvl 2 and 3 can performed automatedly and result be the model
The text was updated successfully, but these errors were encountered:
Farazioo
changed the title
How To Train Tesseract On Custom Fonts Of a Language!
Required GUI Based App To Train Data On fonts And Different Languages
Jun 2, 2024
while trained data is so much different
So As i see the process is like
a picture > box data and ground data > other stuff > trained model
What i think process should be like
lvl 1
--- Input ---
lvl 2
--- Training Data Generation ---
These Steps can be automated
lvl 3
--- Final Output ---
User inputs using lvl 1 and lvl 2 and 3 can performed automatedly and result be the model
The text was updated successfully, but these errors were encountered: