Edit

Point cloud rendering

  • Article

Note

The ARR point cloud rendering feature is currently in public preview.

This feature is being actively developed, and may not be complete. It's made available on a “Preview” basis. You can test and use this feature in your scenarios, and provide feedback.

For more information, see Supplemental Terms of Use for Microsoft Azure Previews.

ARR supports rendering of point clouds as an alternative to triangular meshes. Point cloud rendering enables new use cases where converting point clouds to triangular meshes as a preprocessing step is either impractical (turnaround times, complexity) or if the conversion process drops important detail.

Similar to triangular mesh conversion, point cloud conversion doesn't decimate the input data.

Point cloud conversion

Conversion of point cloud assets works fully analogously to converting triangular meshes: A single point cloud input file is converted to an .arrAsset file, which in turn can be consumed by the runtime API for loading.

The list of supported point cloud file formats can be found in the model conversion section.

Conversion settings specifically for point cloud files are explained in the conversion settings paragraph.

Size limitations

Point cloud asset conversion has a hard limit of 12.5 billion points per converted asset. If larger data sets need to be rendered, the source file needs to be split into multiple assets that obey the 12.5 billion points constraint each. The renderer doesn't limit you on the number of unique assets being loaded, and the streaming data technique ensures that prioritization works seamlessly across all loaded instances. For the overall maximum number of allowed points loaded and rendered by ARR, the same kind of distinctions between a standard and premium rendering session applies, as described in paragraph about server size limits.

Global rendering properties

There's a single API to access global rendering settings for point clouds. The _Experimental suffix has been added to indicate that the API is currently in public preview and might be subject to change.

void ChangeGlobalPointCloudSettings(RenderingSession session)
{
    PointCloudSettings settings = session.Connection.PointCloudSettings_Experimental;

    // Make all points bigger (default = 1.0)
    settings.PointSizeScale = 1.25f;
}

void ChangeGlobalPointCloudSettings(ApiHandle<RenderingSession> session)
{
    ApiHandle<PointCloudSettings> settings = session->Connection()->PointCloudSettings_Experimental();

    // Make all points bigger (default = 1.0)
    settings->SetPointSizeScale(1.25f);
}

Point cloud data streaming

Point cloud asset files are automatically configured for dynamic data streaming during conversion. That means that unlike triangular mesh assets, point cloud assets of significant size aren't fully downloaded to the rendering VM, but rather partially loaded from storage as needed.

Regardless of the point cloud file size, the great benefit of the data streaming approach is that the renderer can start early with presenting the data. The decision of the renderer which data to prioritize, is based on camera view and proximity across all loaded point cloud models. No custom interaction through the API is necessary. Furthermore, data streaming automatically manages the budget and priorities based on how much particular data is relevant for the current view. In case multiple point cloud assets are instantiated on the scene, the streaming system makes sure to prioritize data seamlessly across all point clouds, just as it would be a single asset. Accordingly, splitting the source file is a convenient way to work around the size limitation per file.

API documentation

Next steps