@@ninjafruitchilled If it would generate meshes it would be no different from photogrammetry which then would massively increase the CPU overhead and be no longer an advantage over the "classical" approach.
I was literally thinking of this last night, Gaussian Splatting always breaks up when you look at it from the wrong angle, and as arduous as the process of capturing every aspect of the environment is, it's far easier than modeling by hand, I was thinking about attaching sensors to the camera and capturing color and depth information as well, and the A.I analysing the footage and applying material properties and PBR functionality to the materials in the captured scenes. just spitballing here, it would definitely revolutionize CGI.
Forgot I have x1.25 speed by default and almost died. Also, mind-blowing research and super promising! Would be cool if the "splat to mesh via AI" thing becomes usable.
we have to keep in mind that we don't need the best graphics to actually revolutionize game graphics, imagine combining this with ai generated minecraft worlds or pixel styled games, idk
Google said at one point they would mix satellite and street view together for photogrammetry but i dont know what happened with that because they have never done it yet. that would be crazy for Google Earth VR because the streets wont looks like blobs.
Still waiting for the implementation code on this one. Checkout SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering
