Culling bugs (or missing features) ? about osgjs HOT 12 CLOSED

radius is not transformed ? (means we assume a uniform scale transformation here ? should be said in comments at least. )

You are rigth, that's a bug, radius should be recalculated, not only the bs center.

aren't we forgetting big object intersecting multiple planes case ?

Yep, but it works for the vast majority of the cases, not sure if it is worth to do the extratests. Do you have an example of the culling not working ok with a Big Object?

from osgjs.

Ok @Kuranes could you add push a test that fails, it should be good to fix this to avoid surprise when working with culling

from osgjs.

Added on my todo list.

from osgjs.

@Kuranes Ok, I think I understand it now, we are not culling objects that maybe could be culled. For that we can add a second culling pass using BoundingBoxes, not sure if it is better to do it or not (obviously writting some comments about that possible false positive). Thoughts?

from osgjs.

My guess is that it's really better to have it.
But it would indeed make much more sense with a spatial partition tree (octree/kdtree cull much faster if doing it like that), or for special nodes like paging terrain, so maybe that could be reserved for those types of node. ( hierarchical node scene graph should benefits too, but not as much as bbox are typically more tight)

from osgjs.

Paging terrain is something that I will need to address sooner than later,
for now it works well with pointcloudviz. I'm using an regular octree
so upper nodes have less vertex than bottom ones. That also probably will
be my use case for terrains, but I might see if it becomes an overhead.
For now I don't think it is a blocker. In my opinion a comment should be
enough for the next release. @cedricpinson

El miércoles, 15 de octubre de 2014, Tuan Kuranes [email protected]
escribió:

My guess is that it's really better to have it.
But it would indeed make much more sense with a spatial partition tree
(octree/kdtree cull much faster if doing it like that), or for special
nodes like paging terrain, so maybe that could be reserved for those types
of node. ( hierarchical node scene graph should benefits too, but not as
much as bbox are typically more tight)

—
Reply to this email directly or view it on GitHub
#149 (comment).

Jordi Torres

from osgjs.

I am not sure to follow if it's a bug or if it's an improvement. How it is
implemented in osg?
On Oct 15, 2014 5:10 PM, "Tuan Kuranes" [email protected] wrote:

My guess is that it's really better to have it.
But it would indeed make much more sense with a spatial partition tree
(octree/kdtree cull much faster if doing it like that), or for special
nodes like paging terrain, so maybe that could be reserved for those types
of node. ( hierarchical node scene graph should benefits too, but not as
much as bbox are typically more tight)

—
Reply to this email directly or view it on GitHub
#149 (comment).

from osgjs.

Well it's not a bug, as we are not missing any node to be painted. We are
culling less than we can.

El miércoles, 15 de octubre de 2014, Cedric Pinson [email protected]
escribió:

I am not sure to follow if it's a bug or if it's an improvement. How it is
implemented in osg?
On Oct 15, 2014 5:10 PM, "Tuan Kuranes" <[email protected]
javascript:_e(%7B%7D,'cvml','[email protected]');> wrote:

My guess is that it's really better to have it.
But it would indeed make much more sense with a spatial partition tree
(octree/kdtree cull much faster if doing it like that), or for special
nodes like paging terrain, so maybe that could be reserved for those
types
of node. ( hierarchical node scene graph should benefits too, but not as
much as bbox are typically more tight)

—
Reply to this email directly or view it on GitHub
#149 (comment).

—
Reply to this email directly or view it on GitHub
#149 (comment).

Jordi Torres

from osgjs.

-> radius transformation is a bug
-> culling not all possible cullable not a blocking visible bug, just less precise. Worst case is shadow mapping when you try to get the smallest possible frustum for shadow map precision and might end with a quite bigger one...

from osgjs.

Agreed, that's what I meant. I'm fixing the radius transformation stuff just now.

from osgjs.

@Kuranes #150 should fix the transformation issue, could you test at your end?

from osgjs.

Thanks

from osgjs.