Why is the bounding box received by accept(computeBounds) of a gltf node different from the bounding box in Blender?

[dov]

I'm wondering whether this is a bug or whether I'm missing something.

I have created a model in Blender and found that its size is 579,283,44. I then save it as a glb, load it into a vsg::Node by vsgXchange and read_cast() and then call accept(computeBounds) on the node.

The bounding box received by computeBounds is:

min = [-73.045 -183.394 -344.824]
max = [573.043 462.693 301.264]

Why the mismatch?

[robertosfield]

Unless someone comes forward with an explanation of how Blender computes the size there really isn't much we can say.

You have your data, the source code to the VSG and vsgXchange so can walk through all the elements in play so are in the best position to investigate.

[xiaoxin0214]

Perhaps the reason is related to useNodeBounds. You could try setting it to false and see if it resolves the issue.

        /// Use the bounding volumes of Cull/LOD nodes etc to compute the bounds rather than traverse their subgraphs.
        /// Using the bounding volumes is faster but may result in less tight bounds around the geometry in the scene.
        bool useNodeBounds = true;

[dov]

Thanks! Indeed, turning of the useNodeBounds solved the issue.

I investigated this a bit more and found that the gltf loader creates a cull node which is the bounding sphere of the object. And when useNodeBounds is true for computeBounds,, we get the bounding box of the bounding sphere, which is larger than the original bounding box.

It would be nice if computeBounds would return the original bbox, instead of the bbox of the bounding sphere... But I see that in vsgXchange/src/gltf/SceneGraphBuilder.cpp the bbox is thrown away after calculating the bounding sphere..

In any case, I can easily calculate and cache it on my own.