Is node instancing valid?

[vinjn]

I got some collda files that heavily use <library_nodes >and <instance_geometry>.

The converted gltf looks like this, unimportant details are skipped:

"nodes": {
    // reusable node
    "base0": {
        "children": [],
        "meshes": [
            "mesh0"
        ],
    },
    "base1": {
        "children": [],
        "meshes": [
            "mesh1"
        ],
    },

    // nodes that use base0
    "node_101": {
        "children": [
            "base0"
        ],
    },
    "node_101": {
        "children": [
            "base0"
        ],
    },

    // nodes that use base1
    "node_10": {
        "children": [
            "base1"
        ],
    },
    "node_100": {
        "children": [
            "base1"
        ],
    },
}

However, neither Three.js nor Cesium can correctly render the mesh, only base0 and base1 are visible.

Is this the limitation of gltf format or bug of Three.js and Cesium?

[fabrobinet]

I don't think this:

 "node_101": {
        "children": [
            "base0"
        ],
    },
    "node_101": {
        "children": [
            "base0"
        ],
    },

can happen.

Anyhow, i'd like to have a look, can you send me a test model ? thank you.

[vinjn]

Dae file uploaded as https://cloud.githubusercontent.com/assets/558657/3093329/694702ee-e5b1-11e3-9c48-138430dd89c6.jpg
You can download the jpg, and rename the extension to 7z and extract it.

Thank you!

[fabrobinet]

Ok thanks, I just checked, I am not exactly sure about the problem you are reporting.

1. First thing is it seems you use an older version of the converter (I don't find same ids for mesh).
2. I suspect many leaves should have in the DAE the double_sided extra flag to get expected rendering.

That's the rendering I get because of 2. foliage does not appear as dense as it could be (but overall looks correct ?):

You can see that if I rotate the camera I can see the faces discarded by culling:

You can verify with this viewer:
https://github.com/fabrobinet/glTF-webgl-viewer/blob/master/README.md

However, neither Three.js nor Cesium can correctly render the mesh, only base0 and base1 are visible.
Is this the limitation of gltf format or bug of Three.js and Cesium?

We need first to figure out if something is wrong in the generated file. Also, is there a a viewer that can display just the DAE file as you expect ? just to be sure the COLLADA is matches your expectations.

[fabrobinet]

Well, I feel like I have some documentation to add in the viewer just pointed out above :)...
So if you'd like to visualize your model, just convert it , move it in the /model/folder and add an entry in stage.js, here: https://github.com/fabrobinet/glTF-webgl-viewer/blob/master/ui/stage.reel/stage.js#L92-L96

[vinjn]

Thanks @fabrobinet, your result looks pretty good.
It could be the problem of old converter tool.
I will try new converter and your viewer now.

[vinjn]

All right, I can have the correct result from glTF-webgl-viewer.

However, the result rendered by Three.js is wrong.

It seems it's the bug of Cesium and Three.js, they happen to lack the support for instanced node.
I will report bugs in their issues.

[fabrobinet]

Please check this out @tparisi @pjcozzi thanks.

[pjcozzi]

@vinjn you are welcome to submit an issue to Cesium and we'll look at it eventually. Please provide the test model and a link to this issue.