This repository is an upload of an old project I created in 2020. It is not actively maintained and was mostly a playground for experimenting with neural networks and automation for one of my favorite games at the time.
Code and experiments for automating game agents, including tools for dataset creation, model training, and agent control.
I found an old video of the system working. It was excecuted on a very small virtual machine, so the video is slightly grainy.
- Automates game agents using various neural network models and scripts.
- Includes tools for dataset creation, model training, and agent control.
- Designed for experimentation and rapid prototyping.
game-agents/
├── main.py # Main bot logic and entry point
├── README.md # Project documentation
├── user.cfg # Stores user configuration via pickling (e.g., click locations)
├── model_development/ # Model and dataset scripts
│ ├── create_nonbinary_dataset.py
│ ├── create_nonbinary_model.py
│ ├── model_binary_v1.py
│ └── data/ # Was used to store old game images; avoiding for copyright
├── models/ # Expected location for neural network models
To run a network of bots, I set up virtual machines on Proxmox, all hosted on a refurbished server in my garage. This allowed for distributed agent control and large-scale testing without cloud costs.
Apologies for the code quality—this project was mostly experimental, and I was learning about neural networks as I went.