About visualiazation with SIBR viewers. How does it generate the 3d model. What files are important in the process. about gaussian-splatting HOT 2 OPEN

Those files are important only for SIBRviewer. It need them for initialization. In the practice if you provide anything valid (even with wrong cameras), it will start, and you probably only need to find your target.

from gaussian-splatting.

Those files are important only for SIBRviewer. It need them for initialization. In the practice if you provide anything valid (even with wrong cameras), it will start, and you probably only need to find your target.

Much appreciate!! I found that the cfg_args file may decide what camera SIBR use to build a 3d model.
For example, I copied the local model from 00000 and use the global model to replace the point cloud folder like this:

Then I use SIBR to build a model and it tells me it uses 137 cameras and 137 images:

Also, it mentioned that it used the Colmap file which corresponds to the colmap of local model 00000.
It generates a model:

Then I use SIBR to build the real 00000 local model

and got:


Both of them use the same number of camera so I guess the cameras might be the same. And the COLMAP file they applied are the same too. But the 3d model in SIBR is a little bit different (probably be affect by the model). So I guess maybe I should generate a COLMAP file for the Global model itself? Otherwise I can not get the full, real 3d model of Global Model?

from gaussian-splatting.