You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
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~
The text was updated successfully, but these errors were encountered:
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~
The text was updated successfully, but these errors were encountered: