Hello everyone,

When I train models by LSTM 4.0 
https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#fine-tuning-for-%C2%B1-a-few-characters
 
<[https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#fine-tuning-for-%C2%B1-a-few-characters]>

It showed this problem, how to solve it? Why happened? What should the pic 
size be?


Image too large to learn!! Size = 2608x48
Image not trainable
Image too large to learn!! Size = 2845x48
Image not trainable
At iteration 18938/35100/36699, Mean rms=0.379%, delta=0.441%, char train=
1.315%, word train=2.366%, skip ratio=3.2%,  New worst char error = 1.315 
wrote checkpoint.

Image too large to learn!! Size = 2704x48
Image not trainable
Image too large to learn!! Size = 2666x48
Image not trainable
Image too large to learn!! Size = 2634x48
Image not trainable
Image too large to learn!! Size = 2593x48
Image not trainable
Image too large to learn!! Size = 2593x48
Image not trainable
At iteration 18971/35200/36804, Mean rms=0.376%, delta=0.435%, char train=
1.279%, word train=2.311%, skip ratio=3.4%,  New worst char error = 1.279 
wrote checkpoint.



The pics is created by the shell "training/tesstrain.sh ", How to decide 
the size of the pic by this way???


training/tesstrain.sh  \
   --fonts_dir /usr/share/fonts \
  --lang chi_sim  \
  --linedata_only \
  --noextract_font_properties \
  --exposures "0"    \
  --langdata_dir ../langdata \
  --tessdata_dir /home/hs/tessdata/tessdata  \
  --output_dir ../tesstutorial/chi_sim



-- 
You received this message because you are subscribed to the Google Groups 
"tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/tesseract-ocr/3c59a5c8-0681-43fa-8a63-f8941d59b285%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to