mgr_.Init(traineddata_path.c_str()):Error:Assert failed:

[iuriigalaida]

I've got error 'mgr_.Init(traineddata_path.c_str()):Error:Assert failed:in file ../../../../lstm/lstmtrainer.h, line 110' when trying to run a command below. Can somebody help me with this?
All paths are correct.
version 4.0

c:\Temp\tesseract-master\tesseract-master\api>lstmtraining --model_output C:/Temp/engbest/model --continue_from C:/Temp/engbest/eng.lstm --traineddata C:/Temp/engbest/eng.traineddata --old_traineddata C:/Temp/tesseract-master/tesseract-master/tessdata/eng.traineddata --train_listfile C:/Temp/engbest/eng.training_files.txt --max_iterations 3600

[stephenyong2005]

Do you put the traineddate in the write directory?
--traineddata C:/Temp/engbest/eng.traineddata
Please check that if there is eng.traineddata in C:/Temp/engbest/? If no, copy the eng.traineddata to C:/Temp/engbest/.

Then, I think this will be resolved. I have met the same problem, and this is my resolution.

[iuriigalaida]

It helped. Thanks. But than I got

Loaded file C:/Temp/engbest/eng.lstm, unpacking...
Warning: LSTMTrainer deserialized an LSTMRecognizer!
Code range changed from 105 to 105!!
Failed to continue from: C:/Temp/engbest/eng.lstm

stephenyong2005 Did you see error like this in your case?

[stephenyong2005]

Sorry, I do not got such a error like you mentioned, but i got another error. By the way, I am a new user of tesseract. Do you know how to generate a lstm file? Thanks. 2017年8月13日 16:09，iuriigalaida <notifications@github.com>写道： It helped. Thanks. But than I got Loaded file C:/Temp/engbest/eng.lstm, unpacking... Warning: LSTMTrainer deserialized an LSTMRecognizer! Code range changed from 105 to 105!! Failed to continue from: C:/Temp/engbest/eng.lstm stephenyong2005 Did you see error like this in your case? — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#1075 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AJBNQGx7L5w2ALkdkGuuWtd3SzwqXgI2ks5sX1fwgaJpZM4O0iiS>.

[Shreeshrii]

Are you using the latest source from github for building tesseract? What is your version info, git log info?

[Shreeshrii]

Also see #1069
regarding 'Failed to continue from' error

[stephenyong2005]

I use v4.0.0. Compile from the source code instead of git. OS is ubuntu 17.04. I have finished the demo procedure in the wiki web page of training tesseract 4.0 by following the tutorial. 2017年8月14日 00:18，Shreeshrii <notifications@github.com>写道： Are you using the latest source from github for building tesseract? What is your version info, git log info? — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#1075 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AJBNQPaOqYXe8oSSfc6LX8aVtkQ-yNRIks5sX8qFgaJpZM4O0iiS>.

[stephenyong2005]

this is the tesseract wiki homepage: https://github.com/tesseract-ocr/tesseract/wiki 2017年8月14日 00:56，Shreeshrii <notifications@github.com>写道：

Compile from the source code instead of git.

from where? ShreeDevi

____________________________________________________________ भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com

On Mon, Aug 14, 2017 at 9:57 AM, stephenyong2005 ***@***.***> wrote: I use v4.0.0. Compile from the source code instead of git. OS is ubuntu 17.04. I have finished the demo procedure in the wiki web page of training tesseract 4.0 by following the tutorial. 2017年8月14日 00:18，Shreeshrii ***@***.***>写道： Are you using the latest source from github for building tesseract? What is your version info, git log info? — You are receiving this because you commented. Reply to this email directly, view it on GitHub<https://github.com/ tesseract-ocr/tesseract#1075#issuecomment-322096544>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ AJBNQPaOqYXe8oSSfc6LX8aVtkQ-yNRIks5sX8qFgaJpZM4O0iiS>. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1075 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AE2_ozzaFTAJzfednKsAdNPo1yuTOjnJks5sX8yZgaJpZM4O0iiS> .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub<#1075 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AJBNQGHjjAab3USEZ0D4HZKlj1RaZ8f_ks5sX9NqgaJpZM4O0iiS>.

[iuriigalaida]

After I get latest sourses v4 lstmtraining started work but I almost always got errors like
Encoding of string failed!
Can't encode transcription:
Even if I directly set path to unicharset file.

[Shreeshrii]

That means you have new characters in your training text, so fine tuning may not work.

…

On 14-Aug-2017 4:08 PM, "iuriigalaida" ***@***.***> wrote: After I get latest sourses v4 lstmtraining started work but I almost always got errors like Encoding of string failed! Can't encode transcription: Even if I directly set path to unicharset file. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1075 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AE2_ox6hWfOOzKYP7lTVLo-hmT6EzaRFks5sYCOsgaJpZM4O0iiS> .

[iuriigalaida]

yes. i have 44 new character and I wont to extend english traineddata. Attached my unicharset file. How can I aviod this encoding issues?

[Shreeshrii]

Unicharset was not attached -

see https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#training-just-a-few-layers

Have you tried Latin.traineddata - it will have more characters than eng in it.

[iuriigalaida]

eng.zip

[Shreeshrii]

Thanks for the unicharset. I notice it has accented english letters. 1. Test with Latin.traineddata - which is for Latin script not Latin language (lat). 2. Check that all your additional characters are in best/Latin.traineddata - use combine_tessdata -u to unpack the traineddata and look at its unicharset. I find it easy to sort a copy of the unicharsets and then compare . 3. If Latin.traineddata has all characters you want, then you can do finetune using Latin.traineddata as the continue_from rather than English.

[Shreeshrii]

Re: 'mgr_.Init(traineddata_path.c_str()):Error:Assert failed:in file ../../../../lstm/lstmtrainer.h, line 110'

This comes when the path to the traineddata files is incorrect. Check file locations and correct the command for training.

[amitdo]

The command should use tprintf() with a meaningful message and exit(), not assert.

[Shreeshrii]

@stweil for #1423

[amitdo]

tesseract/src/training/unicharset/lstmtrainer.h

Line 99 in 66b77e6

ASSERT_HOST(mgr_.Init(traineddata_path.c_str()));

[zdenop]

What is the expected solution here? Just note that path does not exist?

diff --git a/src/ccmain/tessedit.cpp b/src/ccmain/tessedit.cpp
index 3ad8d921..6b094129 100644
--- a/src/ccmain/tessedit.cpp
+++ b/src/ccmain/tessedit.cpp
@@ -90,7 +90,6 @@ bool Tesseract::init_tesseract_lang_data(const std::string &arg0,
   // Initialize TessdataManager.
   std::string tessdata_path = language_data_path_prefix + kTrainedDataSuffix;
   if (!mgr->is_loaded() && !mgr->Init(tessdata_path.c_str())) {
-    tprintf("Error opening data file %s\n", tessdata_path.c_str());
     tprintf(
         "Please make sure the TESSDATA_PREFIX environment variable is set"
         " to your \"tessdata\" directory.\n");
diff --git a/src/ccutil/serialis.cpp b/src/ccutil/serialis.cpp
index d9c9a8d4..4356d027 100644
--- a/src/ccutil/serialis.cpp
+++ b/src/ccutil/serialis.cpp
@@ -21,6 +21,7 @@
 #include "errcode.h"

 #include "helpers.h" // for ReverseN
+#include "tprintf.h" // for tprintf

 #include <climits> // for INT_MAX
 #include <cstdio>
@@ -44,6 +45,8 @@ bool LoadDataFromFile(const char *filename, std::vector<char> *data) {
       result = static_cast<long>(fread(&(*data)[0], 1, size, fp)) == size;
     }
     fclose(fp);
+  } else {
+    tprintf("Error opening data file '%s'!\n", filename);
   }
   return result;
 }
diff --git a/src/ccutil/tessdatamanager.cpp b/src/ccutil/tessdatamanager.cpp
index 279cf7ac..1582ef92 100644
--- a/src/ccutil/tessdatamanager.cpp
+++ b/src/ccutil/tessdatamanager.cpp
@@ -82,6 +82,8 @@ bool TessdataManager::LoadArchiveFile(const char *filename) {
       result = is_loaded_;
     }
     archive_read_free(a);
+  } else {
+    tprintf("Error opening data file '%s'!\n", filename);
   }
   return result;
 }

[stweil]

Fixed in commit 68017db.