[Help wanted]Image-recognition-based bot interface[in-progress][0.0]

[rarkenin]

Hello all,

Recently, I had decided to make a bot for Minetest, run from the client-side. Now, I am preparing to ramp up development, now that I have hand-tested an algorithm to read off a scene in Minetest and process it to get node information, as well as designing on paper a system to most efficiently perform a task. The system uses images from the original program:



However, OS-specific interactions such as sending keys and reliably getting screenshots are, to be frank, beyond my understanding at a level where my code would be reliable. Therefore, I am seeking a few programmers that are good with system interfacing and have a bit of time, to help. Since the processing algorithms, so far, are still being designed on paper and pseudocode, my language selection is flexible, most likely C++ with possible Python for some system interfacing, and a PHP configuration script. I'm also planning to use MySQL (or possibly SQLite) to store quite a bit of data, so PHP will pair nicely with it. If anyone wants to help, criticize, give an idea, or ridicule me, please do so.

The image detection in a nutshell:

Detect red by HSV. Build gridlines. Map texels of each face to an image pixel. Check these texels to identify node identities and direction. Angle of lines between red points helps determine direction, also.

I'll also post the texture pack as needed.

Edit: Superseded by http://forum.minetest.net/viewtopic.php?id=4964

[OmniStudent]

So are planning to get the graphics from the client and then process it with image recognition?

[rarkenin]

So are planning to get the graphics from the client and then process it with image recognition?

Indeed, that is so.

[OmniStudent]

That's pretty cool, and also the complete opposite of what I'm doing:

I'm trying to get RID of all the graphics in my bot program, while you pretty much only want to use the graphics!

[rarkenin]

That's pretty cool, and also the complete opposite of what I'm doing:

I'm trying to get RID of all the graphics in my bot program, while you pretty much only want to use the graphics!

Indeed. I need to obsessively rid the viewport of any and all text when the keyboard isn't sending a virtual command. Do you know how to interface with the system to get screenshots? The window titles are bugging me on Windows(as is not having a Windows machine, I cross-compile), and Xorg is just being a pain. If so, can you help with that?
I have an algorithm ready in pseudocode, and tested to work.