Hello everyone,
My team and I recently had an issue with the Google photorealistic tileset not loading correctly around Lille in France, you can make the test yourself using the Sandcaste at this link: Cesium Sandcastle and with this address: “212 rue Colbert Lille“.
At a certain zoom level and tilt the tiles start unloading or are being masked by something, you can also see it in the video below.
That’s a strange one. This looks like actual geometry that is displayed there for whatever reason (you can see slight variations in the color of that green-ish “wall”…).
It took me a moment to find it (but found it by matching the video). Here, a camera position where it is visible is captured:
I then looked around and tried to find the “end” of this “thing”, and … I found it:
It strongly looks like there are some tiles with a waaaay to low LOD are randomly popping in:
(Note: It will be difficult to say with absolute certainly whether this is a data issue, or an issue in CesiumJS. It’s interesting in any case…)
Hi Marco,
thanks for the reply. I’m working with Anthelme, as a PO in his team.
In the meantime, I was in touch with @lisabos, who told me to tag her regarding the issue.
@lisabos above is the replication of the issue we encountered, with a reply from @Marco13.
How can we work together to solve this ?
Hello - Thank you for the details, and thanks to @Marco13 for the investigation. We’ll work with our colleagues on the CesiumJS team as well as report this to Google.
I have a hunch that this might be related to (or similar to) Wrong tiles selected for refinement in certain areas · Issue #12683 · CesiumGS/cesium · GitHub - just to have the cross-link.
Hello everyone ?
Regarding my last message, do you have any clue on how to solve this problem ? @Marco13 @lisabos
We have users who are affected by the issue and would like to find a solution.
Hello Cesium team,
Following up after 28 days: we’re still seeing tile artifacts/disappearances with Google Photorealistic 3D Tiles at certain zoom/tilt angles. This is a blocking issue for our users in production.
Do you have any update on the diagnosis, and is there a workaround you recommend in the meantime?
Thanks in advance!
Just responding because I did some of the initial investigation and I was tagged: I’m not aware of any further updates. It is not entirely clear whether this is an issue with CesiumJS, or an issue with the underlying data. I suspect that it is the latter. One approach for collecting evidence here could be to check this particular region with a different 3D Tiles renderer.
Another instance of this issue (in another location) has been reported in this post. It’s unclear to me whether this is a data issue (Google) or a renderer issue (Cesium).