Hi everyone,
Our engineering team is currently developing a lightweight, containerized UI sharding engine designed to process real-time transaction overlays on live sports broadcasts.
We have our baseline WebGL/HTML5 canvas layer deployed via Google Cloud Run here:
๐ SpatialGridOS Active Demo Stream
As you can see in the demo link, the system maps up to 255 viewer-navigable coordinate nodes directly downstream on top of the live broadcast frame. It handles all interaction logic and transaction processing independently to prevent server infrastructure drag on the macro environment.
Our next objective: We are pivoting to integrate the NVIDIA Holoscan for Media SDK directly upstream to handle automated AI content segmentation, player tagging, and real-time metadata generation. We want to take the metadata outputs from the Holoscan pipeline and use them to dynamically update the coordinate points on our active WebGL canvas layer.
Our question for the Solutions Architects here: Given that our canvas is cloud-native and rendering downstream in a standard browser viewport, what is the recommended practice for piping the frame-accurate metadata payload out of a Holoscan-managed container into an external WebGL coordinate API with minimal latency?
Are there specific reference architectures or GitHub repositories utilizing the Media2 stack or Holoscan metadata operators that we should clone as a starting foundation?
Appreciate any technical guidance or direction you can share!