Item information extraction (need custom ui peice)

[Jida]

First off, Hello!

My name is John and I < collect data for an items site >

I want to increase the usefulness of my website and i am running into a road block.

I currently am trying to figure out to extract item information (stats) from within the game. I am using djpeg and gocr to accomplish this feat.

When i take something like



and run "djpeg -bmp -grayscale -colors 2 test.jpg test.bmp"

This means to change it to greyscale and to 2 colors, it ends up being something like this



You see my issue ... no OCR can read that .... (I can't use the color one in its current form due to it gives the OCR a hemorrhage)

Is it possible for a interface creator to create a small "mod" that changes ITEM links to have the following

A white background
black text for everything.
using standard fonts (or a way to change the font so i can muck with it till i get it working)
no images (if possible)


This will allow me to speed up the data extraction phase and serve everyone with updated fast information =)

If anyone has suggestions on how to extract information from item links using a different method, please let me know
________
MOTORCYCLE TIRES

[Deathbane27]

This will work for items, but not skills. Will finish later, time for Star Trek.

Edit: Updated. Should work for items, spells, and effects.

[lordebon]

I do something like that for recipes (and will eventually do items) only I dump the data to a /say stream and then use a special PHP-based parser to extract the data.

Never thought of using OCR frankly, how well does it work for you?

[Jida]

As soon as i get home tomorrow I will take a look at this and see if the OCR feature work with the updated screen shot XML file.

Valentines day and all ;-)

What i will do is to make a C# program that connects to my MYSQL database and will generate a link biased on information in the tables. Then paste that into EQ2 via the standard game interface.

It will then take a screen shot and auto crop.

Once auto cropped it will be placed through the ocr using the djpeg and gocr programs i spoke about earlier (search Google for the terms if you want to know more).

the text scanned from the image will be parsed and placed into the web database.



IF i can get everything working correctly, the links will be auto validated and link information will be put into the website automatically =).

So, I HOPE that the XML uploaded by the generous developer works as I intended. =)
________
Anime Hentai

[Jida]

Well.. heres what I got.

The problems that I came up with are



Looks ok, but when changed to monochrome to go through the parser.



The issue is the lighter text is not transforming to B/W text.

Also, is there a way to space out the characters? I need more spacing to ensure the OCR reads the output properly. Also, a bigger font would give the ocr better pattern recognition.

The output at the moment looks like this

T_nspDks an allv tD their hD_ vilI_e.

mis spell _ui_ an Odygv StDne.

Ta_et       Raid orGmup Fhend
Pmer       LZlB
Ca_ing      30.O xDnds
R_st       3 minut_
Range       UptDLO.O_en
Level       L3

mde_      Odindion



TelepDm ta_et tD a _one land_h
_u must be a _ti_en tD benetit t_m an vd_y.
mis mAt wnnd be cast duhng combd
It nd a Cih _mer


R_en_

I really appreciate the help

John
________
Pov blowjob

[Deathbane27]

Hmm... Not sure if we can do anything about the lighter text, it's hard-coded to come up that color. Might be able to script around it. Edit: Nope, nothing I can do about it. Only option would be to make it white text on a black background, since that's closer to what the game is expecting when it's picking colors.

Here's with the font enlarged to 20 pt bold Arial.

[gm9]

If you need to do OCR instead of just parsing the text from the logs then you should get an OCR program that does not require B/W images. You generally get much better results if you OCR greyscale (or even color). I never had good results trying B/W scans because characters are fuzzy with the extreme contrast.

Otherwise I would agree, make it white on black, if necessary for grey/yellow text increase the contrast a bit and then invert the image (you OCR should have that option anyway).

[Jida]

Is there a way to extract the data from item link "text" without OCR?

even if it dumped that info into the log, i could use it.....
________
Mexico hotels

[Deathbane27]

Made a button that should /say all of the text on the item examine page. This may or may not work. Can't test, account is currently inactive (I take a week off between months of play to give myself a break. And to let Vitality restore to full. )

[Jida]

Cant wait till i get home
________
Vaporizer save money

[gm9]

It should work. We can also tag the different sections to make them easier to parse, i.e. [DESC]...[/DESC] or something like that.

[Deathbane27]

Well, I tried it myself and it doesn't work. :/ Out of ideas here.

[Jida]

Well.. I had to put it on hold for a few days while I work on my home computer.

Do I just change the #0000000 inside the file to change the colors?
________
Percocet rehab dicussion

[Jida]

The one that changes the font crashes the game. do i have to turn on custom fonts?
________
AMATEUR PORN