Due to computing limitations some of the photogrammetric models I'm creating I need to subdivide the large model space into smaller sub-sections which I can locally process.
These are very high resolution models, but the computer runs out of memory if I don't sub-divide the space into manageable chunks.
With the model space ending up being sub-divided into a grid, and each grid area gets a new 3d model. All the vertex coordinates for all the models use the global origin, meaning that if all the models were loaded into a separate mesh editing software, they would load up in the correct position.
When I export a tiled model, the global origin of my models gets lost. Each model has its origin in the center of the local model space.
How can retain the information of that global origin, so that when multiple tiled models (loading multiple tileset.json files) will load up in the correct relative locations in Cesium viewer?
Can I edit the tileset.json file after the generation of the tiled LoD model is exported?
If so, what would I change to add that relative location of each model to one another?
Are you tiling it with cesium.com/ion online or with the 3D Tiles CLI?
When you say “the computer runs out of memory”, do you mean the CLI tool runs out of memory during tiling? If so, how big is the original high resolution model?
If you’re using the hosted tools, uploading all the sub-divided models together should keep their relative positioning. With the CLI, if you pass the same --longitude/–latitude while tiling the individual models, they should all end up having the same global origin when loaded in the Cesium viewer.
Let me know if that helps.
I don't think i was clear in my original message.
My computer runs out of memory in the photogrammetry steps, so I have to break up the model into smaller sub-sections to generate 3d models from images. This happens when there's tens of thousands of images to process in the photogrammetry step.
The question is how to retain the relative positioning of the models during the tiling step (using CesiumIon online) for all the individual sub-divided models.
Is that even possible?
Thanks for elaborating! This is definitely possible, and I just tested it out to confirm. If you’ve got OBJ files that are positioned relative to each other, if you drag them all into ion and tile them into one tileset, they’ll maintain their relative positioning. This is also a lot more efficient than tiling each OBJ individually since the tiler can make optimizations across the whole thing.
Does that work for you?