nijanthan0

yes, I am using best tess data

lstmtraining --traineddata data/tamtrain/tamtrain.traineddata --old_traineddata tesseract/tessdata/tam.traineddata --continue_from data/tam/tam.lstm --net_spec '[Lfx256 O1c111]' --model_output data/checkpoints --debug_level -1 --learning_rate 20e-4 --train_listfile data/list.train --eval_listfile data/list.eval --max_iterations 3000 Loaded file data/checkpoints_checkpoint, unpacking... Successfully restored trainer from...

I created box file using lsmbox comment and lstmf using lstm.train

Mam but for --eval_listfile I don't know what to give as input so i manually created one impact_condensed font file and then stored in eval listfile. this is my files....

No mam "Compute CTC targets failed!" is not related to impact_condensed eval file.

But I am not using text file in the training process.

Can we directly use the wordstr box file for training?

Mam, thank you for your help.But i have one problem. lstmtraining --traineddata data/tamiltest/tamiltest.traineddata --old_traineddata tesseract/tessdata/tam.traineddata --continue_from data/tam/tam.lstm --perfect_sample_delay 0 --target_error_rate 0.01 --model_output data/checkpoints --debug_level -1 --train_listfile data/list.train --eval_listfile data/list.eval --max_iterations...

How do i reduce the error rate ??

lstmtraining --traineddata data/tamiltest/tamiltest.traineddata --old_traineddata data/tam/tam.traineddata --continue_from data/tam/tam.lstm --perfect_sample_delay 0 --target_error_rate 0.01 --model_output data/checkpoints --debug_level -1 --train_listfile data/list.train --eval_listfile data/list.eval --max_iterations 200 Loaded file data/tam/tam.lstm, unpacking... Warning: LSTMTrainer deserialized an LSTMRecognizer!...