psm argument doesn't work with Tesseract 4.0

[serhii-eleks]

Hello,

I'm trying to launch your environment with tleyden5iwx/open-ocr-2 image.
This image should contain Tesseract 4.0. Looks like decoding image/pdf using psm argument doesn't work.

Request Body:
{
"img_url": "http://bit.ly/ocrimage",
"engine": "tesseract",
"engine_args": {
"config_vars": {
"tessedit_char_whitelist": "0123456789"
},
"psm": "3"
}
}

Reponse:
Error processing image url: . Error: exit status 1

In Tesseract 3.* psm argument use one "-psm", in Tesseract 4.0 two "--psm". I think this is the main issue.

By the way, can you create one addition argument where I can control the output. Not only raw text. I want to receive text in *.hocr format too. And any other. I would be very appreciate to have this feature!

Thanks!

[serhii-eleks]

I found config argument: "tessedit_create_hocr": "1" in order to return data in hocr format.

[tleyden]

Do the docs need to be improved?

[serhii-eleks]

Yes, would be great to improve docs available output formats.
But the main issue is the system cannot execute tesseract with "--psm" parameter.

[serhii-eleks]

Hello @tleyden

Do you have any updates?

Thanks.

[mirko0x5f]

Hello! We're having the exact same problem. We would like to launch tesseract with the psm:3 parameter but we fail to do so for tesseract 4.0.

[mirko0x5f]

The problem seems to be in this line

open-ocr/tesseract_engine.go

Line 87 in 1cd43c1

result = append(result, "-psm")

we managed to fix it only for tesseract 4.0 by changing it in
result = append(result, "--psm")
probably it's needed to switch between the 2 cases to make the change backward compatible.

[overwerk]

we managed to fix it only for tesseract 4.0 by changing it in
result = append(result, "--psm")

Could you please say where exactly I need to replace it? I went into the docker container of the worker and httpd and located the mentioned file, changed it and restarted both containers. Error is still the same.

[thliew]

There is a workaround solution for this issue..
Get into OCR worker container
You should be able to list your running docker container with command below
docker ps
List should contain dockercompose_openocr_1, dockercompose_openocrworker_1, dockercompose_strokewidthtransform_1 and dockercompose_rabbitmq_1 in NAMES column which corresponding to HTTP handler, OCR worker, pre-processor and RabbitMQ.

Use command below to get into OCR worker container
docker exec -it <container_id> /bin/bash

Refactor the source code
cd /opt/go/src/github.com/tleyden/open-ocr/
vim tesseract_engine.go
Around line 87:
Change result = append(result, "-psm") into result = append(result, "--psm")

Recompiling execution file
cd /opt/go/src/github.com/tleyden/open-ocr/cli-worker && go build -v -o open-ocr-worker && cp open-ocr-worker /usr/bin

If you encountered the message below:
cp: cannot create regular file '/usr/bin/open-ocr-worker': Text file busy
you may try to restart the container and try cp it again

Restart docker container
Exit docker container and restart it by using command below:
docker restart <container_id>