there are a lot of flash games that ask u to do repetitive actions. an example would be FarmVille or getting statistics on on-line poker game.
What is the best way to create an interface between what you see on the screen, a bot's algorithm and clicking of the mouse.
so basically, when i'm playing a texas hold'em on face book and if i get AA i want the bot to click all-in. if i don't get AA i want to fold. this is just an example, not the actual strategy. or another examples would be routing tasks in farmville.
so basically. it has to grab a picture of the screen. recognize patterns and convert them into an input of the algorithm. the output algorithms would be clicking on some other (may be same) patterns.
any suggestions?
I have experience building a simple computer vision based bot for a online card game.
I used a crossplatform python library called autopy to handle automated mouse click and keyboard input. The bot capture screen every 0.5 sec, then convert the captured screen into a numpy array for analysis. Python has good image processing (PIL, OpenCV python binding) and machine learning utilities (scikit-learn). For simple image pattern recognition, the bot extract mean of pixel brightness over a image region to make decision. For complicated ones, OpenCV template matching and SVM classifier are being used.
The server caught my bot the first time, so I guess some there's bot detection on the server side. After I added more randomness and flexible decision making, the bot has bypassed serverside bot detection.
All took me a Saturday and I enjoyed Sunday.