What's the difference between neus-facto-angelo and neuralangelo?

[lemondante]

In my test, neus-facto-angelo has better accuracy but sometimes has thin hole on the wall when dataset is indoor scene, while neuralangelo is pretty smooth. I trained with depth and normal.
Are they quite similar? Maybe some regularization or hyper-parameter difference? I check the document and it said neus-facto has proposal network like mip-nerf360.
Btw, I find that resolution of omnidata image is only 384*384, I tried metric3D instead but get quite poor result. Is there any trick to use the depth image? I will try depth-anything v2 later.
Waiting for your reply~