Image to Text ? - Tilted Forum Project Discussion Community

SunGun · 06-29-2004, 08:15 PM

Ok, here's the deal...

I was given thousands of pages of information on a cemetery and they want all this put into a database, and they want it easily changeable. Well I already coded the front end of the database viewer so that can view and change this data after it is inserted into the MySQL database. I just dont feel like typing the thousands of pages into it... They are paying me decently

, but still, it's a computer world, and I know this is possible, so why do somethin that would take me a month the hard way, when i can do it in just a few days with a scanner and a lil enginuity...

Is there a way that I can scan in each *typed* page, and use some program to convert each image into a *.txt or of the like format? That way I can write my own program to parse the information out of there, in a giant loop, and i'll be done with inputting all the information in about as much time as it took to scan each page... It would be so nice

Im hoping this program is free, cuz if it costs too much, I may have to try and write a program that does this myself, and sell it for cheaper... I just know the algorithms I will have to concoct(spelling?) that matches areas of pixelated rigions of an image to a templated pattern to pop out the letter the algorithm thinks this is will have to be damn good, especially to be able to turn someones chicken scratch writing into Times New Roman font.

Thanks for the input
- SunGun

SinisterMotives · 06-29-2004, 08:22 PM

Do a Google search on "character recognition software". You should be able to find something for under $500. Be advised that you will probably still have to proofread the output for the software's "typos".

Moobie · 06-29-2004, 09:26 PM

Most scanner should come with at OCR program bundled with them. That's Optical Character Recognition. They work with varying degrees of reliability so as SinisterMotives said you'll still need to proof read it. They have problems with copying special characters and usually they totally screw the pooch on formatting. But for bulk text transfers it saves so much time and might keep you from getting carpal tunnel.

*edit*
Shit I missed the last line of your post. Are you looking for something that will translate handwriting into type? You may be hosed if that's what you're looking for, cause whatever software you're needing will have to be keyed to the handwriting of that person. Probably impossible, or at very least horribly expensive.

DelayedReaction · 06-29-2004, 10:52 PM

Handwriting is analogous to picture recognition, which is nearly impossible for computers to do. You may be facing a task which would take less time to do by hand.

SunGun · 06-30-2004, 05:07 PM

Im fairly sure 75% of the data is TYPED data... and about the data recognition problems, yeah im going to make my parser read the data into a GUI application, where I can read it, edit any errors, and then hit *save* to teh database, and it will pull in the next persons information that was scanned in, i'll proofread the fields, fix, and save, etc, until im done

That was the plan. A $500 application is too steep, i'll probably try writing my own... thx for the insight fellas

ironman · 07-15-2004, 03:00 PM

get yourself an OCR software, that will do it

06-29-2004, 08:15 PM	#1 (permalink)
SunGun Upright Location: Ravenswood	Image to Text ? Ok, here's the deal... I was given thousands of pages of information on a cemetery and they want all this put into a database, and they want it easily changeable. Well I already coded the front end of the database viewer so that can view and change this data after it is inserted into the MySQL database. I just dont feel like typing the thousands of pages into it... They are paying me decently , but still, it's a computer world, and I know this is possible, so why do somethin that would take me a month the hard way, when i can do it in just a few days with a scanner and a lil enginuity... Is there a way that I can scan in each typed page, and use some program to convert each image into a *.txt or of the like format? That way I can write my own program to parse the information out of there, in a giant loop, and i'll be done with inputting all the information in about as much time as it took to scan each page... It would be so nice Im hoping this program is free, cuz if it costs too much, I may have to try and write a program that does this myself, and sell it for cheaper... I just know the algorithms I will have to concoct(spelling?) that matches areas of pixelated rigions of an image to a templated pattern to pop out the letter the algorithm thinks this is will have to be damn good, especially to be able to turn someones chicken scratch writing into Times New Roman font. Thanks for the input - SunGun __________________ Wow.

06-29-2004, 09:26 PM	#3 (permalink)
Moobie Insane Location: baked beans	Most scanner should come with at OCR program bundled with them. That's Optical Character Recognition. They work with varying degrees of reliability so as SinisterMotives said you'll still need to proof read it. They have problems with copying special characters and usually they totally screw the pooch on formatting. But for bulk text transfers it saves so much time and might keep you from getting carpal tunnel. edit Shit I missed the last line of your post. Are you looking for something that will translate handwriting into type? You may be hosed if that's what you're looking for, cause whatever software you're needing will have to be keyed to the handwriting of that person. Probably impossible, or at very least horribly expensive. __________________ Obscenity is the crutch of inarticulate motherfuckers. We like money. Give us your money you stupid consumer whore. Last edited by Moobie; 06-29-2004 at 09:29 PM..

06-29-2004, 10:52 PM	#4 (permalink)
DelayedReaction BFG Builder Location: University of Maryland	Handwriting is analogous to picture recognition, which is nearly impossible for computers to do. You may be facing a task which would take less time to do by hand. __________________ If ignorance is bliss, you must be having an orgasm.

06-30-2004, 05:07 PM	#5 (permalink)
SunGun Upright Location: Ravenswood	Im fairly sure 75% of the data is TYPED data... and about the data recognition problems, yeah im going to make my parser read the data into a GUI application, where I can read it, edit any errors, and then hit save to teh database, and it will pull in the next persons information that was scanned in, i'll proofread the fields, fix, and save, etc, until im done That was the plan. A $500 application is too steep, i'll probably try writing my own... thx for the insight fellas __________________ Wow.

07-15-2004, 03:00 PM	#6 (permalink)
ironman Smithers, release the hounds Location: Guatemala, Guatemala	get yourself an OCR software, that will do it __________________ If I agreed with you we´d both be wrong

06-29-2004, 08:22 PM	#2 (permalink)
SinisterMotives Junkie	Do a Google search on "character recognition software". You should be able to find something for under $500. Be advised that you will probably still have to proofread the output for the software's "typos".