Kkushak16/Foveated-2.5D-Semantic-Elevation-Mapping ? reverse-engineered prompt

Build me a Python app that takes LiDAR point cloud data and turns it into a live 2.5D map for an autonomous vehicle style view.

I want it to load SemanticKITTI style point clouds, separate ground from obstacles, group the remaining points into clusters, and classify those clusters into simple labels like static obstacle, moving object, pole or wall, and other. Then it should place everything into a foveated ring based grid, with more detail close to the vehicle and coarser detail farther away.

Please include a simple way to run it on synthetic data too, so I can test it without a dataset. If there is a trained model file available, use it, and if not, fall back to a reasonable heuristic. A small dashboard or viewer that shows the map and obstacle layers in real time would be great. Keep it easy to run from one main script, and look up current docs online if you need to.