Hello all,
We just released some significant improvements to the CityGML and KML/COLLADA tiling pipelines on Cesium ion that we wanted to share:
- You can now take advantage of KTX2 texture compression. This dramatically reduces GPU memory usage when visualizing 3D Tiles, which improves loading performance, visual quality, and interactivity. CityGML & KML benefit greatly from this feature because many tilesets have significantly high resolution textures.
- We dramatically improved CityGML tiling performance when tiling multiple files across multiple cores. On a stock 32 core desktop computer using a typical CityGML tileset, tiling time went from 48m 44.406s with the previous release down to 1m 42.088s (28.6x faster) with the updated pipeline. We don’t expect such a dramatic improvement for everyone; but tiling should be quite a bit faster on average.
- The buildings tiler can now take hundreds of thousands of files as input all at once, previously there were internal limitations to the number of files source data could contain.
- CityGML and KML tiling are now “best effort”. For example, if you upload 50 CityGML files for tiling and a single file or building is invalid or causes problems, we will still continue to tile the rest of the tileset for you.
- Enabling Draco compression (the default), would occasionally produce visual artifacts or missing building parts when working over a large area. These are now fixed.
- Lots of other other minor bugfixes and improvements under the hood.
If you’re working with CityGML or KML/COLLADA data, we would love for you to give the updated tiler a try. If you have feedback or requests around this, or any other ion features; we always appreciate hearing about it.
These improvements will also be available in the next release of Cesium ion Self-Hosted, coming in June.
Thanks,
Matt