#26 · Primary category： Game AI & Behavior Trees
NeuralKart
A Real-time Mario Kart 64 AI using ConvNets.
Project last updated：12/21/21
GitHub Stars
671
Forks
56
Contributors
3
License
Other
Why we included this project
NeuralKart is a research demo, not a product: a Keras/TensorFlow model reads the screen through the BizHawk emulator's Lua scripting and steers Mario Kart 64 in real time. An offline search AI generates the training recordings, and a DAGGER-style loop keeps refining the policy. The codebase is small enough to read end to end, and the README walks through running a live demo, retraining weights on fresh recordings, and restarting the improvement cycle. Setup is Windows-only and there are rough edges, but pre-trained weights and save states are ready to download, so you can watch it drive three tracks without training anything yourself. For anyone studying imitation learning or vision-based game agents, the class writeup and explanation video are worth reading for the background.
Articles for this project
No articles for this project yet.
To suggest a topic or contribute an article, contact us.
Related projects in this category
recastnavigation
Industry-standard navigation-mesh toolset for games
mage
XMage - Magic Another Game Engine
lc0
Open source neural network chess engine with GPU acceleration and broad hardware support.
beehave
🐝 behavior tree AI for Godot Engine
forge
An unofficial rules engine for the world's greatest card game.