Large WebGL

[carmine arch]

It streams the data in at two parts, one is asset bundle data chunks (large 20x20km of 3D models cubed into 1m sizes with LODs) and the other is file data reading (potree for a larger 30x30km point cloud overlay) with the initial load being only about 70mb. The hard part there though is you can really only load about 500mb of data safely at most. It was a lot of work, I think it took us a good 7 months to build that particular project (it was for an engineering company). It needed an entire pipeline.

So they drop files in a drive, and a server process splits the images, videos, models and point cloud files into different pathways for processing (i.e. FME & blender for the models, potree for the point cloud, etc) and then triggers a custom unity build (with the target companies branding, icons, screenshots, and data). It then spits out projects for each platform and begins deploying. It takes up to 10 hours end to end depending on how much data is sent. The biggest data test was 100gb.

What we do is load and unload asset bundles based on where they are - so it removes unused memory and old regions as you go. This lets us bring in 500m+ tris of 3d data no problem.

Potree lets us do small data loads per region and does really good quality up close and less quality in the distance. The best part of that one is the octree file serving system, which lets you handle billions of points of data easily.