Sorry for not posting anything lately, I’ve been too busy to find topics with programing some image recognition algorithms.
Here is a clip of what I’ve been doing lately:
working example:
Sorry for not posting anything lately, I’ve been too busy to find topics with programing some image recognition algorithms.
Here is a clip of what I’ve been doing lately:
working example:
You must be logged in to post a comment.
I like what your doing. I can see lots of room for improvement too. I noticed that you loop from left to right and top to bottom on comparing pixels. So I assume you must search the screen capture in a similar manner. If speed is an issue, I would suggest searching from the center of the source image outward or even try checking pixels in half way iterations in a divide and conquer algorithm. I don’t see much on the Internet for techniques in monitoring a game but I would presume it would be reasonable to have the program save the typical destination of where the searched image is found.
I play this game frequently and I would really like to see your program come to fruition. Send me an e-mail with the source and we’ll talk!
searching from the inside to out would actually greatly increase the required time, since it would be adding a lot more value checks.
right now it searches the screenshot for an image in less than 10ms, while using less than 10% cpu, which is better than all of the current .net based image searching libraries. Most of them take 1+ seconds.
your assuming that i’m some amateur programmer who can’t program worth diddley squat.
So you must be using a Boyer-Moore type algorithm to scan in one direction? Computers are much faster now-a-days anyway but image searching isn’t quite there for fractal image compression to be practical. I rewrote a Windows screen saver I made in pure Assembly language to get it to a reasonable screen rate.
I like the blog. Ideas intrinsically have value. I was getting my hair cut and the barber asked what I did and I told him I was making cell phone apps at the time. Impressed as he was, I turned it on him and asked, “what kind of a novel app do you think is needed” and he was dumbstruck. So the lesson is that you really need to be keen on finding something useful and desirable.
There’s a niche that your blog is filling and I have some ideas that would work off of what I see. Besides playing with 3D algorithms in Processing.org lately, in the past month I had come up with how I could apply Artificial Intelligence algorithms to a game involving lots of economical decisions. If there is still a need for speed for your program, I’ve thought of a more involved solution for image scanning.
There’s always potential in someone who is willing to engineer around a problem and work out a program. The higher level of mathematics that are involved in a line of work, the greater the pay is. Send me an e-mail. I’m still interested in seeing your program too.
Hi there. I am learning c# and was wondering if you might consider selling your image recognition / detection code in c# for me to use and learn from? I would love to hear from you. Cheers, Tim.
yeah sure… whats your offer?
AU$25 …. and another $25 once it’s working as you have it on your youtube videos. So AU$50 all up ….. quite a reasonable offer I would imagine.
I am just new to C# but I have been working on some image detection code but it’s just not fast enough. I saw your videos working on the flash games and your code seems pretty responsive. Any idea how long it takes to detect an image using your code ?
Let me know what you think.
Cheers !
Tim
not really… considering that its the only decent .net based image recognition library out there…
and the last person i let get a hold of it started up his own bots, which directly compete with me, making me lost $600 a month.
No troubles. Unfortunately I can’t really justify more than that because it is just for my own use and own learning – I have no commercial interests in making bots to sell to others.
Anyway, I have learnt a lot since I made the first post and now have my image recognition down to <200ms for a full 1920×1080 screen. This is a long way off the 5-6 seconds it was taking me with my earlier rounds of code.
I would still be interested if you have any advice for me, as I respect the skills you have.
If you don't mind putting me on the right track can you answer this one question ? How do you incorporate a % tolerance into your image recognition methods?
Cheers !
well i guess if its just for you i can get it to you for $50.
add me on skype: hotnoob.