ImageMaze

ImageMaze

This demo reads a PNG and draws every non-white pixel as a coloured cube. The white pixels become the paths through the maze and the red troll can only move onto those pixels.

There are OpenGL and WebGPU versions. They share the image coordinates, collision tests and actor state in maze_scene.py, so both versions behave in the same way.

Running

uv run --script main.py
uv run --script main_webgpu.py

The original small.png map is used by default. The other two maps from the C++ demo can be selected from the command line.

uv run --script main.py --map maps/colour.png
uv run --script main_webgpu.py --map maps/test.png

Controls

• Arrow keys move the troll through white pixels.
• 1 selects the overhead camera.
• 2 selects the troll camera.
• W toggles the maze wireframe.
• Left-drag rotates, right-drag pans and the wheel zooms.
• Space resets the mouse transform.
• Escape quits.

The WebGPU wireframe is built from explicit cube edges. This avoids depending on the optional WebGPU polygon-line feature whilst keeping the same control as the OpenGL version.