Training for Urdu
36 views
Skip to first unread message
Samee Arif
Nov 14, 2021, 2:13:44 PM11/14/21
to tesseract-ocr
Hello!
I am training LSTM based tesseract for Urdu language. I have 85 .tif images and each image has 5000 Urdu text lines. So total text lines are 425,000. Are 200,000 iterations with 0.001 learning rate, [1,48,0,1Ct3,3,16Mp3,3Lfys64Lfx96Lrx96Lfx512O1c1] network specs, 0.5 momentum and 0.999 adam beta enough to get good results?
Thankyou
Thankyou
Reply all
Reply to author
Forward
0 new messages