Impossible Game AI

A neural network trained to beat the impossible game

How does it work

This network learns from examples of people playing the game, so we need to record the data beforehand. Because we dont want out network to learn from people's mistakes, we remove automatically the deaths fragments in the original videos. Once all this is done, and after some light preprocesing, the data is packed in a compressed numpy file, which is used to train the model.

We can test our model by running the game and letting the neural network decide when It has to jump.

Do you need powerful hardware to run the network

Yes, but if the neural network lacks speed, you can always try reducing the speed of the game using Cheat Engine

Reference

Video

This project is based on: https://github.com/ikergarcia1996/Self-Driving-Car-in-Video-Games

Name		Name	Last commit message	Last commit date
Latest commit History 43 Commits
.idea		.idea
images		images
keras		keras
src		src
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
ScreenRecorder.py		ScreenRecorder.py
grabber.py		grabber.py
keyboard_controller.py		keyboard_controller.py
process_game.py		process_game.py
record_data.py		record_data.py
run_game.py		run_game.py
videoClipDeaths.py		videoClipDeaths.py
view_video.py		view_video.py

License

druskus20/allmighty_ai

Folders and files

Latest commit

History

Repository files navigation

Impossible Game AI

How does it work

Do you need powerful hardware to run the network

Reference

About

Resources

License

Stars

Watchers

Forks

Languages