Do you need more information about my installation and my training command ? Or should i just evaluate using the a--trainedata ?
thanks in advance. Le jeudi 22 juin 2023 à 14:38:51 UTC+2, Augustin Fourcaud a écrit : > hello, i finetuned the eng.traineddata model because i wanted it to > reconize the greek lambda symbol. I got the ocrlambda.traineddata file and > i want to evaluate it using lstmeval. > > when i eval a checkpoint file with --trainedata parameter set to > eng.traineddata i get terrible results with this error on every iteration > where the lambda appear. > ************************ > Encoding of string failed! Failure bytes: ce bb 4d 4e 44 > Can't encode transcription: 'LY O kcXλMND' in language '' > Truth:8Az7V I vUOs > OCR :i g8 A z. 7 V I vlU O0 s.l > Line BCER=1.000000, BWER=0.666667 > ************************* > > but when i train with a --trainedata parameter set to > ocrlambda.traineddata i get really good results. > > but on the doc a saw that the trainedata parameter should be set to the > file > that was given to the trainer. > > is it an error or do i understand or do anything wrong ? > > thanks a lot > -- 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 tesseract-ocr+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/e37e1900-96c2-4dd4-91e5-a0b3c6502a70n%40googlegroups.com.
