Indoor scene semantic segmentation

Thank you for this excellent project!
I am trying to use your model to perform semantic segmentation on my own dataset. My dataset is in 3D point cloud format(.ply)(I can also convert it to other formats), containing xyz coordinates, RGB values, and normal vectors. I am eager to apply your model for semantic segmentation of my indoor scene. It is similar to the s3dis dataset, but it does not contain any label information. My goal is to segment walls, ceilings, and floors through your model.
Could you kindly provide a simple demo or guide me on how to utilize the excellent Sonata model for this task? Thank you！

No file, test, or entry point is named in the issue. Begin by locating Sonata's existing inference and dataset-usage guidance, then determine how its model handles an unlabeled .ply point cloud containing xyz, RGB, and normals; done means a simple guide or demo explains segmentation of indoor walls, ceilings, and floors on data similar to S3DIS.

Written by the indexing model from the issue text.