Developed an AI bot in a video game with image recognition ability using python language.
What does the AI do: The AI recognizes specific pattern when it appears on screen and distinguishes them from among others and performs the desired task.
How I did it: I had to input around 800 sample images of each distinct pattern through which the AI learnt and created a model respectively. I used the TensorFlow library for this task. Once the AI recognized the image correctly, I made it perform the desired actions with basic python coding.
Few of the libraries used: numpy, tensorflow, PIL, pyautogui, imagesearch etc
The entire task of the bot was to farm in a game while recognizing the different elements that appear on the screen. Although the accuracy of the AI is not 100% (around 99.7%), it has never failed so far. I had a really mind-boggling yet fun experience creating this one.