PDA

View Full Version : OCR from photos, Maemo 5 on N900, just like SonyEricsson P1i


illemann
2010-02-13, 21:44
hi,

i would like to see OCR support on N900. I used it alot on my SE P1i. Be able to take a picture on a business card or to grab an article or recipe, time table, pricelist and convert it to text is just so powerful and usable!


what do you say?

Apoc
2010-02-13, 21:54
+1 with support for tables so I could easily grab my work schedule and stick it on my server for my other lazy co-workers :D

illemann
2010-02-14, 08:53
lets not get to hasty here :)...dont know if there exist any OCR that can read tables, especially if we are talking open source. Though there is one open source software that could be of help here:
http://en.wikipedia.org/wiki/Tesseract_(software)

In this case its just almost about making the GUI thats missing and of course JPG implementation.

Bec
2010-02-18, 13:45
Here's a brainstorm for it, some progress has been made: http://maemo.org/community/brainstorm/view/ocr_for_n900

RevdKathy
2010-02-19, 08:40
I would imagine this should be possible this app (http://www.youtube.com/watch?v=adPdZzMlm3M) is doing something similar and then taking it a step further,. i would assume.

dwould
2010-03-01, 22:45
so i was playing today and I compiled tesseract for n900. i also compiled ImageMagic so i could convert from jpg to tif for tesseract.

providing your picture is black text on white backround it seems to work pretty well.

i'm wondering how hard it would be to write a 'sharing' plugin for the photo manager, which will just call a script to process the image through convert and tesseract and spit out the text.

if i get the time I'll play more. might be cool if someone with a setup for packaging would consider uploading tesseract and magemagic to extras-devel....

quipper8
2010-03-01, 23:13
so i was playing today and I compiled tesseract for n900. i also compiled ImageMagic so i could convert from jpg to tif for tesseract.

providing your picture is black text on white backround it seems to work pretty well.

i'm wondering how hard it would be to write a 'sharing' plugin for the photo manager, which will just call a script to process the image through convert and tesseract and spit out the text.

if i get the time I'll play more. might be cool if someone with a setup for packaging would consider uploading tesseract and magemagic to extras-devel....

I was gonna do the same thing, but did not get around to it yet.
I also use unpaper (http://unpaper.berlios.de/) when automatically processing scans on server.

it is a small program but helps a good bit with tesseract ocr

i tie together some imagemagick->unpaper->tesseract with pretty great results and insert the raw text output into db as metadata for searching docs on our server.

works pretty well.

dwould
2010-03-02, 07:49
thanks for the tip. i will check it out today.

dwould
2010-03-02, 08:30
Sadly unpaper compiles fine, but segfaults when run under the ARMEL target. appears to be fine under the X86 target. I think I'm probably at the limits of my understanding to attempt to figure out what it doesn't like.

Anyone that understands C what to take a look?