I was looking for "NeRF" in here but reading the paper it seems like they don't use neural networks to represent anything.<p>They seem to just optimize the positions and shapes of gaussian primitives as well as the reflectance properties.<p>Certainly a lot more "explainable" than a NeRF.