Solver for Sigmar's Garden

This is a dumb, hacky, WIP solver for Sigmar's Garden which is the mini-game in Opus Magnum.

It works in a few steps:

1. Take a screenshot of the game.
2. Build a model of the game state using image classification.
3. Solve the game with a simple depth-first search.
4. Click buttons to actually execute the solution.
5. Rinse and repeat.

Requirements

• Python 3.6
• This runs on Windows only at the moment. (It uses win32 APIs directly).

Initial setup

You'll need to train the image classifier before getting started.

1. Start Opus Magnum and open up Sigmar's Garden.
2. python -m sigmar.vision.training generate
3. Manually categorize the generated images into the provided folders.
4. python -m sigmar.vision.training train

Thanks

The image classification code is heavily based on this blog post by Vikas Gupta and code from this Gist by Stephen Meschke. A million thanks to them for making it all click for me finally.