Optical caracter recognition: gocr to be used with sgrab output.

	Bottom Previous Topic Next Topic
Register To Post

JosDuchIt

Posted on: 2013/3/10 16:48 #1

Just can't stay away

Has anybody experience with gocr ? it needs gzip bzip22 & netpbm and system should support popen(3) (??) to be able to use eg jpg files that sgrab can produce.
gzip seems only available in 68 k version, a "similar" one, named dact is availmable in OS4depot. Can it be used?

Chris

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/10 18:30 #2

Amigans Defender

You mean this?

It's rubbish, but it works.
(I suspect newer versions are better though)

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/10 19:20 #3

Just can't stay away

I downloaded it from OS4Depot, but it is the same version (0.46)

Anything better as OCR for the Amiga ?

salass00

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/12 9:15 #4

Just can't stay away

@JosDuchIt

I just compiled gocr 0.50 which is the currently latest version. You could try it if it works better for you:

https://dl.dropbox.com/u/26599983/gocr-0.50.7z

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/12 14:19 #5

Just can't stay away

@salass00 thanks
@chris can you tell me how you installed it (see questions in 1st mail) ( and maybe test the new compile?)

Where oe all thes files of gzip bzip2 netpbm files go?
can i mix up th 68k gzip files with the rest?
Can dact (OS4 replacement?) be used instead of gzip 68k

.

Edited by JosDuchIt on 2013/3/12 14:53:49

Chris

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/12 18:52 #6

Amigans Defender

@JosDuchIt

I'm afraid it's so long ago I can't remember. If it's just calling the command line versions (which it sounds like from your description), then simply putting them in the path should do the trick.

NetPBM is quite big so I'd advise chucking it on its own somewhere, and putting "path work:netpbm add" into user-startup.

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/13 11:00 #7

Just can't stay away

@Chris
I tested the new compile with 2 "excellent" .pbm files indicated on the gocr website

I got crashes in both cases. (it might well be that it crashes on the previous compile , i did not test this)

5.Datas:Graphics/gocr> gocr -i font1.pbm -o ram:tks2
file downloaded from
http://jocr.sourceforge.net/tmp/examples/clean/font1.pbm.gz

5.Datas:Graphics/gocr> gocr -i font2.pbm -o ram:tks2
file downloaded from
http://jocr.sourceforge.net/tmp/examples/clean/font2.pbm.gz

pvanni

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/13 14:20 #8

Not too shy to talk

If you haven't tried to increase the stack, do it
try with a very big one f.e. 1000000

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/13 17:00 #9

Just can't stay away

@pvanni
Thanks that helped

Ihave now been trying with .jpg files(text grabbed with SGrab) and get this

Quote:

8.Datas:Graphics/gocr> gocr -i Snoopy.jpg -o Snoopy.txt

read-PNM-error: file number is 3, position -1
read-PNM-error: bad magic bytes, expect 0x50 0x3[1-6] but got 0x64 0x6a8.

The textfile Snoopy.txt is empty, and i don't see anything created in the /db folder.
I was at least expecting something there generated from the successful OCR read of the .pbm files.

Snoopy does not give me FAIL's that i can link to a bad installation.

BTW Bzip2 and gzip were present in the OS4 SDK and the netpbm/bin is in the path too.

broadblues

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/14 9:48 #10

Home away from home

Quote:

NetPBM is quite big so I'd advise chucking it on its own somewhere, and putting "path work:netpbm add" into user-startup.

path work:netpbm/bin/ add

Blender For OS4.x : Blues : Walker Broad

broadblues

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/14 9:51 #11

Home away from home

Quote:

The textfile Snoopy.txt is empty, and i don't see anything created in the /db folder.
I was at least expecting something there generated from the successful OCR read of the .pbm files.

Snoopy does not give me FAIL's that i can link to a bad installation.

BTW Bzip2 and gzip were present in the OS4 SDK and the netpbm/bin is in the path too.

You need to convert the jpgs to PNM first. That's what you need netpbm for.

gocr doesn't do this for you.

[edit]
On scanning the docs it appears that gocr needs pnmtools not netpbm. If you only have netpbm you will need to do the conversion to pnm yourself.
[/edit]

Blender For OS4.x : Blues : Walker Broad

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/14 12:15 #12

Just can't stay away

@broadblues Thanks

I did find pnmtools.c here:
http://src.gnu-darwin.org/ports/print ... k/pnm2ppa-1.12/pnmtools.c
and pnmtools.h here:
http://pnm2ppa.sourcearchive.com/docu ... 5/pnmtools_8h-source.html

pnmtools.c has no main though. Maybe it should be compiled with gocr??

xenic

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/15 0:58 #13

Just can't stay away

I just grabbed a text window with SGrab, converted the ilbm image to a ppm with dttoppm and used gocr to convert it to text. It works but has some errors in the converted text. Is there any way to train gocr for new fonts?

Wasn't gocr the evil Babylonian god in Ghostbusters?

Amiga X1000 with 2GB memory & OS 4.1FE + Radeon HD 5450

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/15 7:17 #14

Just can't stay away

@xenic
I did use jpgtopnm for converting jpg filesand got lots of errors.

from http://sourceforge.net/projects/jocr/ ... forum/22204/topic/1280843

i then used.
gocr -p ./db/ -m 130 Haldmn50.pnm

You get the possibility to correct errors, but till now i did not succeed in saving them tot db/db.lst

I allways get an errror message (just before the big first characters to correc:

DB /db/db.lst not found
it is quite interesting to be able to do ocr in scripts (eg after having graabbed a number of textparts from a document image), so i hope to find out how to do it properly.

Some basic info here too
http://sid.ethz.ch/debian/gocr/gocr-0.45/doc/gocr.htm

lQuote:

I just grabbed a text window with SGrab, converted the ilbm image to a ppm with dttoppm and used gocr to convert it to text. It works but has some errors in the converted text. Is there any way to train gocr for new fonts?

I did not find dttoppm, i guess it should be ilbmtoppm
The results are not better though using an ilbm output for SGrab.

Edited by JosDuchIt on 2013/3/15 7:49:51

xenic

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/15 18:45 #15

Just can't stay away

@JosDuchIt

I got dttoppm from Os4Depot. It will convert ILBM. It's probably better to save (or scan) the text image to a lossless image format like ILBM or PNG. JPEG doesn't preserve a 100% accurate image.

You might use Snoopy to see where gocr is looking for /db/db.lst.

Amiga X1000 with 2GB memory & OS 4.1FE + Radeon HD 5450

JosDuchIt

Re: Optical caracter recognition: gocr to be used with sgrab output.

Posted on: 2013/3/16 10:47 #16

Just can't stay away

@xenic
Quote:

I got dttoppm from Os4Depot. It will convert ILBM. It's probably better to save (or scan) the text image to a lossless image format like ILBM or PNG. JPEG doesn't preserve a 100% accurate image.

Yes i came to that conclusion too, but i did find a conversion program from ilbm to pnm in netpbm/bin so i used that

SOLVED the problem stupid confusion about paths used in the command line

Edited by JosDuchIt on 2013/3/16 11:07:31
Edited by JosDuchIt on 2013/3/16 11:09:02
Edited by JosDuchIt on 2013/3/16 11:18:49

Register To Post
	Top Previous Topic Next Topic

Currently Active Users Viewing This Thread: 1 ( 0 members and 1 Anonymous Users )