Re: Adria - Inventory Management System
Let's see here ...
1) The video is meant to give a rought idea of how to get you going on using the program, until we are 100% sure we're not going to change things around we'll probably not make anything more detailed (unless one of our users wishes to help).
2) The itemscan should be reading your item and marking each unidentified symbol it finds for you to manually identify. Once your library of symbols is complete there should only be a bit of "crap symbols" around the edges that still get detected, these should be left at the default of "?". Once you lock your symbol library (singlelock in this case) adria will stop asking for you to identify new symbols and just assume they are "crap symbols" or otherwise spaces it should ignore.
3) When you identify a symbol/letter you should just imput the symbol/letter and press enter when prompted to do so while scanning an items description. For symbol/images as in 1x1 items while you use multiscan you should input its ID, for runes it's their names (lowercase) and for gems it's their name and a number skull5 would be a perfect skull and skull1 would be a chipped skull. You can find the others and add your own in the OCRMultiScanIDs.txt file you have in your datafolder. (multiscan video shows this process)
For the singlescan/itemscan to correctly categorize scanned items you need to have the text colors somewhat correct (check the program options OCR/color tab). The RGB value listed there should be taken from the brightest pixel of a letter of the specific color and the slider att the bottom adds a plusminus intervall around the number to account for possible variations.
Not sure sure if anything of this solves it but nevertheless ... if you're alright with a remote help session using teamviewer we may be able to fix things if up if it doesn't.
New OCR system pro's:
Doesn't rely on you having same graphic settings as me (better overall adaptability)
Easier to correct symbol identification errors for the user (no hardcoded symbols)
Ability to add new symbols to support things I didn't think of (again not hardcoded)*
99.9% correct inside item description (gray i and j can be pixel value identical if background is working against you from what I've noticed)
99.9% correct on the sides of the item description = little to no crap text (I'd say previous was only close to 90-95%)
Fixed fullscreen scan problems so that should work now (glide mode only)
Con's:
You have to ID most symbols yourself when you're starting out, unless you get your buddy to hand you his symbol libs (they're in the datafolder) ...
Hard to tell when your text symbol lib is really complete ... (I'd just scan a few accounts and then assume it is, unlock it and add symbols whenever you notice something is missing)
Fulscreen symbols might be slightly different then windowed ... make sure adrias listed RGB colors are correct and keep the symbol lib unlocked when starting out ...
Yes, never scale the window when you wish to scan something, the image adria gets from the windows API we use needs to be the standard 800x600 or 640x480.
*: For example an annoyance previously was how a gray U could be split up into two parts because the way symbols are detected using vertical lines and thresholds. In the old version you could get "(J" from the hardcoded identifying routine. In this new version it is fine to think of half-U-first and half-U-last as seperate symbols in the symbol library and add a autocorrection that fixes the problem. Example: Identify the first part of the U and make the symbol identifying routine return "[U" then let the second part be returned as "U]" now all gray U's are returned as "[UU]", a rather unique string that you can add an autofix for that replaces it with just "U", without risking replacing any true "(J" with Us.