Theory craft. Making my OWN OCR how to start out?

[butterfly]

OCR = Optical Character Reading

 

Hello all, I am theory crafting here for now and all help is appreciated

First of all I would like to remind everyone that i am not a skilled programmer and I am even a slow and poor learner. so if you explain me something please make sure I can understand, I would really be thankful for that.

 

1. I have a program that allowed browse through a digital online brochure / stock and It always used to work perfectly with pixelchecksum() conversion to database and I got what I wanted

but things have changed negatively for me.

They Have changed the look of the digital brochure.

1. They have made fonts more transparent

2.  They have made the font much much smaller often letters are 7pixels high and 5 pixels width

for example the letter Afflicted looks like this;  keep in mind that this is 1600% increased   this image is originally 84x9 pix which is very small

3. They have added a pesky background which messes my autoit Pixelchecksum beyond useable

So my while project / Program is halted to a stop and I already spend more then 1000 hrs in it I have to change my program

I have done some decent analyzing on the screenshots / digital brochure and came to a conclusion that there is always enough RGB Difference between my font and background

So it seems I need to proper filter it, Train it ( i do not mind spending lotsa time into it)

 

My goal is to able to read text as quick as possible , therefor I am looking into my own personal lightweight OCR

I would like it to able to read a page of 50 items of lets say average of 30/35 words each line

and my goal is to achieve this under 500MS  but i would not weep if it would take 1000 ms

speed is very important to me,

What would be best approach?

making / creating a seperate .DLL or Exe for OCR

and if I make a seperate exe or .DLL how does it transfer the information instantly between .exe (or is that impossible)

it would be best for me  that my autoit.exe does not waist time on trying to OCR, my autoit already;

--------------------------------------------------------------------

handles onscreen actions (mousemovement),

other calculations ,

Writing and retrieving to database,

and interaction with a different program

-------------------------------------------------------------------

for example my theory is as follow

 

 

I make an extra ocr.exe or ocr.dll which either prepares out of itself the data upfront or does it when it is asked (having the ability on the fly and change it when

needed during the process would be extremely nice because I can switch between speed and saving Resources if needed)

question 1 =  I am not sure yet how this can be transferred between ? 

1. does it need to written in text file ,

2. can it be saved in physical memory

3. does the ocr. saves itself in some array and gives it out on request from my autoit.exe

Thank you for reading, I will show in a few hrs my theory how to basic program it for speedyness

Edited May 28, 2014 by butterfly

[Jfish]

I would start by searching the forum for OCR ... you will see many hits.  Also, people may want more information about your project because you could easily be talking about Captcha or something similar.  You might also be violating copyright laws for the online catalog.   I think if you explained exactly what you are trying to do (what website, what content, etc) you might get some more responses.

[butterfly]

I am reading an online catalog which is allowed by OCR, I am not interested in Captcha or something similar and Where would I need to have a captcha that would read over 500 fixed Letters words within 1 second?

I am not breaking any laws or terms, and I Like to stay far away from infractions of this forum also

Just for starters I would like to have a good answer of whether I have to run by an include

a separate .exe or separate .dll

thank you very much

Edited May 28, 2014 by butterfly

[JLogan3o13]

As was stated by yourself in the OP, you have no idea how even to begin this project. So why would you not take the suggestion given to you by Jfish and reading some of the many OCR threads on this forum first to see how others have done it, rather than asking someone to give you a crash course from Day 1?

[BrewManNH]

If it's an online catalog, have you tried reading it with the IE functions?

[Bert]

I'm sure your intent is honorable. Understand however we do have a few concerns.

I am reading an online catalog ...

Who's catalog? Name please.

 

I am not breaking any laws or terms,

Link to TOS please. No offense, but we get slapped around from time to time when certain sites find out we automate stuff or do things they do not like. (for example downloading YouTube videos. Google was not thrilled with us for that.)

I am not interested in Captcha or something similar

I get your position and respect it. Please be aware Captcha breaking can be done by the same means with what you are asking for (the same methods can be used) and that is a touchy subject here in the forum.

----------------

We are happy to try to help but we do have to be careful we do not break any rules.

[junkew]

If speed is important why not write this in c.

[butterfly]

If speed is important why not write this in c.

 

I have never done any programming  outside auto-it ha ha  <3

anyway I am very disappointed in this forums attitude towards certain subjects and the unhelpfulness without giving any Single direction or clue.

I have been searching about this topic already dozen of hours with no luck and the only response I am given here is the bad cop approach

I do not want to criticize anyone here but I do not have anyone in my surroundings I can ask any single thing

....... I guess it helps me mold into something.

the good news is:

This following topic was very useful to me

'?do=embed' frameborder='0' data-embedContent>>

I have encountered it before but I overlooked the good parts of it

Although the start of it was very confusing but the 2nd page helped me out a great deal.

I believe I can finish it in 2 or 3 days in autoit for myself and maybe I can post it here . (although this first draft version will be too slow because of many issues)

Thank you anyway

[Jfish]

Everyone said they would help once you confirmed you were not violating any copyright laws by sharing the URL for the site you are targeting with your OCR script.  Still waiting ...

[Jos]

I have never done any programming  outside auto-it ha ha  <3

anyway I am very disappointed in this forums attitude towards certain subjects and the unhelpfulness without giving any Single direction or clue.

I have been searching about this topic already dozen of hours with no luck and the only response I am given here is the bad cop approach

I do not want to criticize anyone here but I do not have anyone in my surroundings I can ask any single thing

....... I guess it helps me mold into something.

you know this mainly has to do with your expectation and not really with the answers given ...right?