Graphics Reference
In-Depth Information
The rest3d open-source project is an experiment to provide client and server
code to study the issues openly and converge toward a practical specification.
Currently, it includes an XML database system, a nodejs http server, and client
code for a basic UI, WebGL viewer, and model loaders. The system is able to
fetch, parse, and display 3D content for both COLLADA and glTF. Parsing data
and preparing it for WebGL visualization is an order of magnitude (10
)
faster with glTF than COLLADA. glTF becomes significantly better for models
that need a lot of content processing.
A sample application was created to demonstrate using a web browser and
server as a glTF conversion service. This shows that using these technologies, it
is straightforward to create an easy-to-use tool that would always be up to date
with the latest runtime format, and that can keep the content uploaded to be
automatically converted when needed. The URL of the converted content can
be used to be incorporated in the assets. Since it is a cloud service, a model
converted a long time ago can keep the same URL, and it will automatically be
reprocessed if needed so that it is always up to date.
Balancing the number of http requests, the size of each unit of content, and
caching behavior is dependent on the type of application and content. rest3d
aims to provide control over those parameters when running the converter and
additional packaging on the server. One area of work, already available in the
rest3d repo, is a geometry compression technique that can be applied to the
geometry at the end of the COLLADA2GLTF processing pipeline. The client
has the option to ask the server to compress the geometry, which will optimize
the bandwidth and transfer time, at the cost of decompression on the client side.
×
-20
×
2.10 Future Work
As we prepare the glTF spec for ratification, we are still finalizing the format
itself and continuing to build open-source tools.
Compression is a promising area to expand on. At the time of writing this
chapter, two early integrations of compression libraries have been tested with
COLLADA2GLTF:
•
Open3DGC—an ecient TFAN compression [Mammou et al. 09],
•
webgl-loader—a fast and compact mesh compression for WebGL [Wonchun
14].
Figure 2.10 shows early results using Open3DGC. These numbers are promis-
ing, and typically for very dense meshes, the ratio between uncompressed and
compressed data is even better. Our next integration step is to investigate mem-
ory footprint and processing impact while decoding compressed meshes.
