There isn’t one formula for calculating it, but usually it’s some metric based on point density, tile sizes in meters, or any other factors. At runtime the geometric error is translated into screen space error and this is used for determining when to load a tile’s children. So in general a higher geometric error means it will load its children sooner when the camera is zooming into it.

If you are building an octree, a common thing is to do is pick a geometric error for the root and then halve the geometric error for each level down the tree.

There isn't one formula for calculating it, but usually it's some metric based on point density, tile sizes in meters, or any other factors. At runtime the geometric error is translated into screen space error and this is used for determining when to load a tile's children. So in general a higher geometric error means it will load its children sooner when the camera is zooming into it.

If you are building an octree, a common thing is to do is pick a geometric error for the root and then halve the geometric error for each level down the tree.

I've already done a lot of searching.

But I can not understand real meaning of GeometricError.

So, I do not know what value to enter in geometricError when creating tiles.josn file.

Does it depend on the specific of my point cloud data?

please answer about my question.

===

information of my pointcloud is follows..

numbers of points: 725,178
average spacing: 1.5 (m)
center coordinates(x, y, z): 322613.55, 4160270.15, 41.43 (m)
bounding box size(dx, dy, dz): 905, 770, 123 (m)

Hi Sean,

What value would you recommend for root's geometricError?
If you have a set of 3D model files with different density of meshes, how will you set the geometricError for each level of model?

For the root you probably want a value proportional to its size, like maybe it’s area or some other metric. Same for each level below. But try some values out, see how it looks in practice, and then tweak if needed.