Tesseract: Running example hocr command from wiki does not work as expected

[DustinWoods]

I'm having trouble simply running the example command from wiki page on hocr output (https://github.com/tesseract-ocr/tesseract/wiki/Command-Line-Usage#hocr-output)
command:
tesseract --tessdata-dir ./ ./testing/eurotext.png ./testing/eurotext-eng -l eng hocr

Environment

• Tesseract Version: 4.0.0-beta.1 with Leptonica
• Platform: Ubuntu
• Directory: The cwd ./ contains eng.traineddata from tessdata_best

Current Behavior:

eurotext-eng.txt is generated with text and the terminal says Can't open hocr:

read_params_file: Can't open hocr

Expected Behavior:

Should generate eurotext-eng.hocr
(Note: this works as expected if I exclude --tessdata-dir ./)

Suggested Fix:

Perhaps this is an argument parse bug, or maybe there's a new syntax and the wiki needs updating. Or maybe I'm missing something.

Thank you!

[stweil]

Thank you.

The example was wrong for your case. You already found the right solution. I fixed the Wiki.

[DustinWoods]

Thanks @stweil. So does this mean the hocr option cannot be combined with the --tessdata-dir option? Is there a workaround to use these two options together?

[stweil]

The Tesseract command line syntax is a bit confusing. hocr is not an option, but a configfile. --tessdata-dir is an option. Both can be used together, but options must come before configfiles. In your case the given directory did not contain the expected tessdata files.

[amitdo]

The Tesseract command line syntax is a bit confusing

It's terrible...

[DustinWoods]

Ah I see. The wiki does call it a config. So I just need to copy tessdata/configs/hocr into my tessdata directory and it should work. I'll give this a try.

[DustinWoods]

Hmm didn't work. Got this error: read_params_file: parameter not found: enable_new_segsearch

[stweil]

It worked. That's only a warning. You copied an old hocr file. Remove enable_new_segsearch from that file.

[DustinWoods]

Actually the hocr config file does not contain enable_new_segsearch. And there is no generated files txt or hocr/html.

Maybe someone else can reproduce what I'm experiencing:

• Using 4.0.0-beta.1 leptonica-1.75.3
• Create directory with just these 3 files:
  
  
  
  • eng.traineddata (from tessdata_best)
  
  
  • hocr
  
  
  • test-image.png
  
  
• run tesseract --tessdata-dir ./ ./test-image.png ./extract-image-output hocr

The hocr file contains:

tessedit_create_hocr 1
hocr_font_info 0

The output from this is:
read_params_file: parameter not found: enable_new_segsearch
and no new files created in the cwd.

[zdenop]

If you really get error read_params_file: parameter not found: enable_new_segsearch there are 2 possibilities:

1. You are run different command than you post
2. Your tesseract installation does not use eng.traineddata (from tessdata_best) as you stated.

Config parameter enable_new_segsearch must be somewhere specified (e.g. in traineddata or config file).
BTW please use recent tesseract version (and data) when you report issue.

[DustinWoods]

Thanks @zdenop for the help! I was able to get things working normally by doing two things:

• I uninstalled 4.0.0 (this was the latest version for Ubuntu using apt)
• I downloaded and built 4.1.0 from source and installed it
• I re-downloaded eng.traineddata from tessdata_best

The result: hocr file was generated and no warning message.

Thanks again for the help.