(Spanish version here)
This bot employs computer vision techniques, optical character recognition (OCR) and reinforcement learning to autonomously learn playing Geochallenge (one of the most popular Facebook flash games).
This bot employs computer vision techniques, optical character recognition (OCR) and reinforcement learning to autonomously learn playing Geochallenge (one of the most popular Facebook flash games).
This program has been developed in standard C++, it is OpenSource (GPL3) and of course, it can be downloaded for free:
- Windows executable: geochallenge_ai_bot-1.0-win32.zip
- Source code: geochallenge_ai_bot-1.0-src.zip. CMake and OpenCV are required to compile it.
The program uses three data directories:
learndata
: All the bot learns is stored here. This directory is NOT included in the downloads, thus the first few times you use the bot, it will fail a lot, but eventually it will learn and improve. If the game language is changed ("English" is recommended to avoid problems with the OCR), this directory must be deleted to allow it to start from scratch.
tessdata
: Files used by tessract, the OCR engine.
data
: Miscelaneous images needed by the bot.
Those who are starting using OpenCV, or in general to play with computer vision techniques, I strongly encorauge them to glance the sources since there are examples of image correlation, finding patterns, thresholding, channel separation, image scaling, etc...
Enjoy!