Clarification on Reproducing Paper Results and Octree-GS Similarity

[ZachariVaia]

I would like to confirm whether the following mappings (config) are correct for reproducing the results presented in the paper:

Anchor 2DGS: explicit -> 2D -> base_sh2_mean

Anchor 3DGS: explicit -> 3D -> base_sh2_mean

Our 2DGS: explicit -> 2D -> lod_sh2_mean

Our 3DGS: explicit -> 3D -> lod_sh2_mean

Scaffold 2DGS: implicit -> 2D -> base_sh2_mean

Scaffold 3DGS: implicit -> 3D -> base_sh2_mean

Our Scaffold 2DGS: implicit -> 2D -> lod_sh2_mean

Our Scaffold 3DGS: implicit -> 3D -> lod_sh2_mean

Additionally, do the hyperparameters remain the same across these configurations?

Lastly, in the previous GitHub repository, is the octree-gs configuration more closely related to Our Scaffold?

Looking forward to your clarification. Thanks!

[cskrren]

Scaffold 2DGS: implicit -> 2D -> base_rgb_mean

Scaffold 3DGS: implicit -> 3D -> base_rgb_mean

Our Scaffold 2DGS: implicit -> 2D -> lod_rgb_mean

Our Scaffold 3DGS: implicit -> 3D -> lod_rgb_mean

[cskrren]

For the explicit 3DGS/2DGS in the main paper, I use SH3. However, recent experimental results have shown that there is not much difference between the results of SH2 and SH3, and you can refer to it for yourself.