Need Some Ideas Here

[big_daddy]

I'm trying to make a script that will follow one color...too easy, but at the same time run from another color. I just can't think of a way to do it.

[Infinitex0]

Maybe something like mousepos(+5 pixelcheck) just a very pesudo* idea.

[big_daddy]

Maybe something like mousepos(+5 pixelcheck) just a very pesudo* idea.

That somewhat what I had in mind, checking +-5 pixels both x & y for the color I'm running from. The problem is how do I run and follow at the same time?

[Infinitex0]

Maybe And

[Uten]

Does the colors change or are they static? If they change in what manar do they change (iratic, progressive?) How fast does the colors change?

This problem can probably be solved with a simple nevral network written in AutIt, depending on your answers.

[big_daddy]

Does the colors change or are they static?

The color I'm following is static, however the three colors I'm running from start washed out then become solid.

If they change in what manar do they change (iratic, progressive?) How fast does the colors change?

Progressive, within 5 seconds.

[Uten]

The color I'm following is static, however the three colors I'm running from start washed out then become solid.

Progressive, within 5 seconds.

I probably asked the wrong question. I take it the pixels on your canvas does change.

1: You have to follow one color.

2: Pixels of the target color moves around the canvas? (do they leave a trail?)

3: You have to make sure you don't get trapped by pixels your running from (RF)?

4: The RF pixels can they change back to something?

Do I have it right so fare?

Can I see the problem on the net somewhere?

It will probably take me some time (I'm a bit rusty) to create a neural network with a simple fuzzy decision tree but it could be a cool UDF (To all of the purists AI geeks:evil: , I DID SAY COOL not efficient! )

[neogia]

If you're following one color, aren't you inherently running from ALL other colors? If there is one black pixel on the screen, and i'm following black, there could be a million different other colors on the screen, but knowing what those other colors are doesn't change my progression, I'm still going to follow black.

Every time I check the screen for if I'm still on the black pixel, I do something like "First Pixel: Not black; Second Pixel: Not black; Third Pixel: Black; Ok, move right 3 pixels; Restart" I think it might be futile to figure out what the color is, and THEN deciding to run from it, even if your already "running from" it to "run toward" the black.

If I'm off base, then I probably just don't understand your purpose. Try fleshing out your concept, or give a concrete example. Is it a game? Picture editing?

[big_daddy]

Okay, here is an example of where you would need to do this.

Dodging Circles

[Thatsgreat2345]

MULTI THREADING 1337NESSSSSSSSSSSSSSSS

[Uten]

Okay, here is an example of where you would need to do this.

Dodging Circles

hmm, wonder if every one here, is there. It does not respond.

EDIT:

Ok, so in this sample you have circles bouncing around in a square. You want to track the green circle (it is not completely green it is lie a sky of dots inside a circle ) but you have to avoid the reddish circles bouncing around. Calculating the entire screen is considered (lets pretend..) to be to time consuming. The circles are moving with different velocity.

THOUGHTS:

1: Lets make our tracker circular (should it try to define the green circle?). Then it needs a radar made of tentacles (we don't have to calculate all pixels in the boundaries in this case).

2: We have to calculate approximately velocity and direction of the green sky.

3: We have to detect red balls moving in our direction.

4: We will encounter a decision problem when a red ball has speed, velocity and size such that the green sky will be covered entirely. Depending on the rules laid out. If we can have our radar span the entire screen it is no problem. If it is allowed to only span a small part, then we do have a problem. Because how do we locate our target cloud when we have to move away from it.

5: In this sample case our (15x15? or only 0,0) pointer icon has to be protected. We die a horrible death if a red pixel enter this area. (looks like the 0,0 location will destroy us)

6: Something coming from behind (moving approximately in the same direction as the green cloud) is less dangerous than something coming from the front.

CHALANGE:

How many pixels does your radar need to get the highest score in this game? Basically your score should be divided by pixels you read in each calculation.

Damn, I don't have time for this Have to lett the back of the brain work for a while

Edited April 25, 2006 by Uten

[Paulie]

couldn't you make a say 15x15 pixelsearch following green, and then a 40x40 avoiding red with a shade value,

make a variable that if a red is detected, tuns off the green follower, and just avoids red, and do the same vice versa. (i never understood how to make an actual PixelSearch function, the first four #'s throw me off, otherwise i'd try writing one) but you get mouse position and add 15 pix to follow green and 40 avoiding red, then make a switch variable that goes from 0 - 1 depending on whether a red pixel is detected or not.

[big_daddy]

1: Lets make our tracker circular (should it try to define the green circle?). Then it needs a radar made of tentacles (we don't have to calculate all pixels in the boundaries in this case).

I think defining the green circle would be a good idea. That way we know where we want to be if we can.

2: We have to calculate approximately velocity and direction of the green sky.

I've been looking though some java game code trying to get some ideas for this.

3: We have to detect red balls moving in our direction.

What I had in mind was a "shield" that is about +- 10 pixels from our cursor that looks for anything other than white or green.

4: We will encounter a decision problem when a red ball has speed, velocity and size such that the green sky will be covered entirely. Depending on the rules laid out. If we can have our radar span the entire screen it is no problem. If it is allowed to only span a small part, then we do have a problem. Because how do we locate our target cloud when we have to move away from it.

Okay ....I think I understand.

5: In this sample case our (15x15? or only 0,0) pointer icon has to be protected. We die a horrible death if a red pixel enter this area. (looks like the 0,0 location will destroy us)

This is true.

6: Something coming from behind (moving approximately in the same direction as the green cloud) is less dangerous than something coming from the front.

This is also true.

CHALANGE:

How many pixels does your radar need to get the highest score in this game? Basically your score should be divided by pixels you read in each calculation.

Not sure what you are asking here? A good score on the game would be over 1500. However the game is just an example of the problem, if we can solve the problem this could be used in many cases (ie making games).

[Uten]

Not sure what you are asking here? A good score on the game would be over 1500. However the game is just an example of the problem, if we can solve the problem this could be used in many cases (ie making games).

The chalange part was just a note about how the "radar" code should be optimised. Few calculations/pixel checks is "better" than checking the entier screen.

You probably know about the aimbot prof of concept by Simucal. It is sortof what you want but you have to add colision detection.

[LineageIIC3Freak]

Well i'm a noob when it comes to pixel searching and I still have questions about it on another topic, but just to put my two cents in I see that all the opposing pixels are different shades of red so when you do a pixel search you just put in a number for the [shade variation] part of pixelsearch and see how that works but hey i dunno im just guessing... GL