how to extract mesh of brandenburg_gate successfully?

Hi~

Thanks for sharing this great code.

I have tried to adapt extract_mesh.ipynb to extract mesh from learned nerf-w model.
after seen the notes , I first get the bounds of point cloud from `dataset.xyz_world`:
```
x y z min:
[-1.58000625 -1.41236581 -0.89322236]
x y z max:
[ 3.32392681 2.49776899 10.96546297]
```

so I set bound params of this scene like:
```
xmin, xmax = -7, 16 # left/right range -1.2 1.2
ymin, ymax = -7, 16 # forward/backward range
zmin, zmax = -7, 16 # up/down range

```

however, the mesh is unreasonable:

have anyone made it ?

Start with extract_mesh.ipynb and inspect how bounds are derived from dataset.xyz_world and applied to the learned NeRF-W model. Reproduce the Brandenburg Gate extraction with the reported bounds, then compare the resulting mesh with the expected scene geometry to identify whether the bounds or extraction process is responsible.

Written by the indexing model from the issue text.