Model recognition

[ragehunter3]

I looked around and didnt find a tut or release of this so i wanted to ask if anyone could please teach me about how to make a program to recognise specific models in a game.
Sorry again if im posting in the wrong area .. already released etc etc.

Thanks up front.

 

[Zeillinger]

What do you mean by "recognise specific models"? There sure are modeling libraries in various APIs for computer graphics, but the answer depends on what you are actually trying to do... Are you talking about computer vision or am I completely off base?

 

[ragehunter3]

Well em im unsure if i explained right but im trying to build an auto targeting system for a game (aim robot) and i was told id need something to target automaticly models (of characters? Like specific parts of the body i guess im unsure) and then just make it fire the trigger. So id have to start off with making it find the people to target using models but im not sure how too.

 

[Zeillinger]

A good way to start is to search for an OCR library / code snippet written in your language of choice. The concept of OCR is very similar to what you've described but instead of searching for characters you'll need to search for the targets. Creating a database of target samples and of various filters applied to those samples (e.g. high contrast) would also be useful.

 

[Justei]

afaik to accomplish this you could hook the process, then using a hook for d3d see if you can get hold of the model bones, then get the head bones, once you have this you should be able to make it aim for the head, either by getting onscreen coordinates or somehow manipulating the raycasting.

Since i never make cheats for fps game this might work or might be totally ridiculous.

Thats how i would look around though.

Hopefully this helped.

 

[ragehunter3]

Ok so some of that i understood some of that .. em .. less xD.
About the OCR Em.. im not sure i fully understood but if i did correctly the point was not to make a database but it would identify the target according to one specific id / model / whatever its ment to be. I.E i know runescape bots use models so there more accurate without using pixel scanners.

So im pretty sure it should be available for some FPS games o.o thought id learn how to make an aim robot detecting the head of enemies or at least some part of them and then triggering the fire hotkey.

@Justei
Em could you give an example of something you've done before? (If you don't mind it being public) cause ive only managed to make a color based aim robot which isnt much of an aim robot but yet just triggers automaticly. Even thats not 100%. Could you also explain a bit more about using the models to aim at people?

 

[Justei]

Ok so some of that i understood some of that .. em .. less xD.
About the OCR Em.. im not sure i fully understood but if i did correctly the point was not to make a database but it would identify the target according to one specific id / model / whatever its ment to be. I.E i know runescape bots use models so there more accurate without using pixel scanners.

So im pretty sure it should be available for some FPS games o.o thought id learn how to make an aim robot detecting the head of enemies or at least some part of them and then triggering the fire hotkey.

@Justei
Em could you give an example of something you've done before? (If you don't mind it being public) cause ive only managed to make a color based aim robot which isnt much of an aim robot but yet just triggers automaticly. Even thats not 100%. Could you also explain a bit more about using the models to aim at people?

I don't have any examples, I only have knowledge of bones etc because I somewhat work with 3d models etc since I develop games.
And those models of for example humans have "bones" etc, those control the movement of the character etc, now if you somehow can start controlling those by hooking d3d you can do a lot of things, as well as locating them on the screen, thus being able to make your aim point to the bone controlling the head etc.

This might work or not, not sure, never really cared about making aimbots etc.

 

[Zeillinger]

A database will boost recognition accuracy (and speed, in some of the cases) but it isn't necessary if you can efficiently recognize characters by some common feature. I still think that looking at the source of some OCR lib will really help you.

Edit: Justei is also suggesting some useful pieces of advice but the process of creating the model really depends on the game graphics. If the characters are different and complex it might be really difficult to create a unified model (a database can be useful in that case).

 

[ragehunter3]

Needing to search more images would reduce speed no?
Anywho what programing languages would work with this? (What ones would i need to learn to know how to make an aim robot)

 

[Justei]

C++, C#, and a bunch of others, but my guess is C++ is one of the more used languages for cheats in general.

 

[ragehunter3]

So do you know any specific tuts for making an aim robot on C++? A link (if thats allowed here) would be most appricated.

 

[jMerliN]

It's less about model detection and more about texture detection. You figure out which textures are drawn on which models, determine which texture is being applied to a primitive by a checksum, then you use primitive data (vertices, normals, etc) to determine coordinates that you want to aim at.

 

[master_unknown]

Aimbots are build from Direct X.
Try reading direct x tutorials.

You may also want to visit this site as they specializes in making AimBots for FPS games such as COD.
Aimbots.net - The n°1 community for all your gaming needs.