Hello, I'm trying to re-train the chi_sim.traineddata model from scratch for studying.
I use the source data of chi_sim.training_text in the link directory https://github.com/tesseract-ocr/langdata/tree/master/chi_sim to train the model with the command: training/lstmtraining --debug_interval 100 \ --traineddata ~/tesstutorial/trainspecial/chi_sim/chi_sim.traineddata \ --net_spec '[1,48,0,1 Ct3,3,16 Mp3,3 Lfys64 Lfx96 Lrx96 Lfx512 O1c1]' \ --model_output ~/tesstutorial/specialoutput/base --learning_rate 20e-4 \ --train_listfile ~/tesstutorial/trainspecial/chi_sim.training_files.txt \ --eval_listfile ~/tesstutorial/evalspecial/chi_sim.training_files.txt \ --max_iterations 3600 &>~/tesstutorial/specialoutput/basetrain.log The net_spec is same as the official model package (chi_sim.traineddata from the tessdata github). After converting the training model with the lstmtraining --stop_training, a new chi_sim.traineddata model gererated, which is named chi_sim_new.traineddata. And I name the official chi_sim.traineddata as chi_sim.traineddata for distinguishing. Then I pull out all the characters in the two traineddata model. There are 4384 characters in the chi_sim.traineddata, but 2538 characters in the chi_sim_new.traineddata which is generated by me. Why are there different characters in the two models? Does the source data in the chi_sim.training_text haven't updated in time? -- 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/1111e3f0-588b-456f-90bf-a878f20b1f26%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
