I have been unable to generate tiles for 3D data that was generated and exported from Agisoft Metashape on Cesium Ion. I have tried uploading data in different formats(obj, glb, kmz) and obj as a compressed .zip as well. File sizes were 3.4Gb, 4Gb, 10Gb and 20Gb.
In all cases, it retrieves the files, starts tiling, gets stuck at 14% for a day and then fails with the following message:
Something went wrong with this asset. To report this issue, please email support@cesium.com with the asset ID: …
Can you help me understand what is the issue here? I also have a few questions:
What is the maximum possible size for a 3D tile source?
What is the preferred/recommended file format and ion options for 3D photogrammetry data?
Would it be possible to estimate some kind of time to finish number?
I apologize that you are seeing these errors. For the assets that have errored would you be able to provide me with asset ids so that I can take a look internally. We are planning to launch a major update to our photogrammetry model pipeline soon, so providing the asset ids will also help ensure our new tiler properly handles these files.
To answer some of your questions:
There is no known limit on size. The only hard limit from ion is that tiling finish within 24 hours.
Cesium recommends .obj, .glb, or .gltf models for model uploads and these formats are natively supported by our pipeline.
There is no way to easily estimate the time it takes to tile data because it is not directly proportional to the size of the input data and there are many other factors. For example, layout of textures and complexity of the model data can both have a dramatic affect on tiling time.
It should not be, unless you share your Ion Access Token, then people would be able to use your assets depending on the settings you have set for the token.
After investigating the underlying issues the team has identified these to be recurring with our current photogrammetry pipeline. But as mentioned above we are hoping to deploy a new version of the pipeline very soon which should address the issues that you are seeing here. Thank you for your patience!
That is very disappointing to hear Ankit, I was really hoping to work this out quickly as we are in the progress of integrating cesium to our stack for the first time. Do you have any timeframes as to when this update will be available? Is it possible to pause/suspend our paid account until the update arrives since we are not able to generate tiles?
The team is trying to run your assets through our new pipeline internally and should upload them to your account soon. I will reach out to you once they are done uploading them.
Hi, We have been waiting for an update for a long time and we are very concerned about the lack of progress. We request you to escalate this issue and revert/contact us for a call to discuss the next steps.
It’s been over a month and a half since we took a subscription for Cesium ion and till now, we do not have a working solution and are stuck without any updates from the dev side. Please consider resolving this at the earliest and also consider pausing our subscription.
@skydeck I’m incredibly sorry for the lack of reply here.
The issue for most of the assets you shared was that our current generation tiling pipeline does not handle this type of data very well. We are in the process of finalizing the next generation tiling pipeline which was built for better scalability, performance, and visual quality and that has taken a little longer than we expected.
In the meantime, we have processed three of your assets with the new pipeline and uploaded them to your account. These are new assets, since the previous uploads appear to have been deleted. The name of the asset is the previous asset ID you shared: 1801303, 1800196, and 1799882.
Two of the assets, 1806460 and 1805330 appeared to be invalid, trying to even open the source models in Blender produced some odd results.
We absolutely want to support your use case and make sure ion ultimately handles this type of data smoothly, efficiently, and without errors. I can possibly put you in contact directly with someone internally if you are interested in sharing some additional details about your use case and data.
Thanks, and sorry again for the extremely slow reply.