TL;DR of the paper, generated automatically.

Research paper proposing P-MGS, a new way to place Gaussians on meshes for editable splatting.

• Idea: Gaussians placed by how complex the surface looks, not by mesh face count.
• Editing: Gaussians bound to the mesh after training, and move with it when the mesh deforms.
• Results: the abstract claims better novel view quality with fewer Gaussians. No numbers given.