Implicit tiling and discrete LOD's?

General
1

In the 1.0 samples there is a sample with discrete LOD’s see 3d-tiles-samples/1.0/TilesetWithDiscreteLOD at main · CesiumGS/3d-tiles-samples · GitHub . I’m wondering how to achieve this effect with 1.1 Implicit Tiling (using a quadtree for example). Any ideas?

I was looking for some option in the tile semantics (3d-tiles/specification/Metadata/Semantics at main · CesiumGS/3d-tiles · GitHub) but not convinced it’s possible to achieve discrete LOD’s using that.

2

anybody have some suggestion?

3

Just to get a better idea: It sounds like the goal here to have an overall structure for the tileset that is similar to the one discussed in another thread.

The difference would/could then be that there is a content available in each of these 3 levels. This content could then just be the different LOD levels of the sample that you linked to. I think that this should be possible (but would have to try it out - maybe I’m overlooking a caveat here…)

4

in the sample 1 tile is replaced/refined by another tile depending on geometric error (so the tiles span the same region).

In the implicit tiling case there is a predefined quadtree, I don’t think there is a possibility to define that a specific tile should be replaced/refined by another (so on the same level) depending on geometric error?
Only possibility I could think of is create the refined/replacement tile on another level (but that would mean splitting the original tile).

5

If the goal is to basically have a fixed-size box, filled with one model in different levels of detail, then the structure of the sample model should accomplish that goal. Implicit tiling usually implies subdividing the model on the lower levels.

At which point do you see a limitation of the approach that is used in the sample model? Or conversely, which problem are you trying to solve with an implicit tiling approach here?

(One could try to force things, with some quirky workarounds or emulations - like overriding the bounding boxes with semantics, to basically “bloat up” one child to occupy the same space as the parent, but I’m not sure whether that would even have the desired effect. At this point, it might rather make sense to have a UNITREE subdivision scheme in addition to QUADTREE and OCTREE :smiley: )