Training Tesseract: unicharset extractor producing "Bad properties"

[guspo...@gmail.com]

In some recent posts, I've seen people with similar problems as mine, but no answer as how to fix it.  I'm trying to train tesseract to be more accurate with a new font.  When creating the unicharset using unicharset_extractor on my box file:

```

a 32 692 165 958 0 

b 221 734 354 958 0 

c 32 446 165 628 0 

d 221 488 354 628 0 

e 32 275 165 373 0 

f 221 317 277 373 0

```

I get the following output:

```

9 

NULL 0 NULL 0 

Joined 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0     # Joined [4a 6f 69 6e 65 64 ] 

|Broken|0|1 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0        # Broken 

a 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0  # a [61 ] 

b 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0  # b [62 ] 

c 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0  # c [63 ] 

d 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0  # d [64 ] 

e 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0  # e [65 ] 

f 0 0,255,0,255,0,0,0,0,0,0 NULL 0 0 0  # f [66 ]

```

and when i run shapeclustering, if gives a the first few lines of:

```

Bad properties for index 3, char a: 0,255 0,255 0,0 0,0 0,0 

Bad properties for index 4, char b: 0,255 0,

```

It seems that the unicharset_extractor isn't properly parsing the box file.  Some obvious problems with the unicharset file are the "properties" bit mask is 0, the "glyph_metrics" field appears invalid (0,255,0,255,0,0,0,0,0,0), the "script" field should be either "Latin" or "Common", but is NULL, etc.

Anyone have an idea why is is happening?

O/S: Ubuntu 15.10

Tesseract Ver: 3.04

Posts with no simple resolution:

https://github.com/tesseract-ocr/tesseract/issues/139

[Meltem Çetiner]

Hi, Im trying to train as well  and I have the same problem. I got this result :

"P 5 0,255,0,255,0,32767,0,32767,0,32767 NULL 54 0 0 # # P [50 ]A

A 5 0,255,0,255,0,32767,0,32767,0,32767 NULL 38 0 0 # # A [41 ]A

S 5 0,255,0,255,0,32767,0,32767,0,32767 NULL 53 0 0 # # S [53 ]A"

I have the problem with the fields of glyph_metric and script. Is there any idea?