You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi there,
Thanks for this open-source OGC-3Dtiles pnts tiler, pretty fast and customizable CLI! Also pretty great that the 2.0.1 version stores coordinates and transforms in local CRS that is then transformed to EPSG:4978 via only the root transform.
We produce some pretty massive tilesets, a few billion points, of photogrammetric 3d-scans done via ground + drone captures - not purely airborne lidar - and I noticed the tiles have a pretty small extent with maximumScreenSpaceError: 1px in CesiumJS and there are lots of them. Plus Cesium using a depth-first rather than breadth-first approach at loading children tiles makes the problem pretty apparent in our case - just opened an issue here.
I tried enforcing a resolution that matches the points spacing at highest resolution, 0.1mm on some focus areas, to see if the resulting tiles would be bigger, but got the error message
log.Fatal("resolution should be between 1 and 1000 meters")
resolution should be between 1 and 1000 meters
Would it be possible to allow for resolution of 0.0001m (0.1 mm)? Do you think this could solve the described problem, or maybe I should try enforcing a min-points-per-tile of 10k or 20k?
The text was updated successfully, but these errors were encountered:
jo-chemla
changed the title
Allow for resolution below 0.5
Allow for resolution CLI flag below 0.5, eg up to 0.0001
Dec 17, 2024
I am not sure if I understand exaxtly the problem here, if not please correct me.
I can certainly change the resolution flag bounds however I'm not confident this will solve your issues. Setting the resolution value to 1mm means that the coarser LOD will contain at least 1 point per mm3 which is a very high value you probably don't want.
This will certainly reduce the depth of the tree Cesium traverses, but I expect the resulting root pnts/glb files to be very large, kind of defeating the purpose of the streaming format.
If the problem is the size and depth of the tile tree you should jointly use depth and min-points-per-tile flags.
depth controls the number of LODs, allowing you to create shallower or deeper trees of tiles. min-points-per-tile allows to stop the creation of a new LOD for a tile in case there are not enough points, which helps optimize the tree making it less deep where not needed.
resolution instead influences how many points are put at each non-leaf LOD. Lower values mean more points, doubling at every level of detail. Leaf nodes always contain all remaining points so resolution doesn't really play a role there.
Without knowing the details of your point clouds (eg extents in meters and number of points) and also more specific description of the issues you are facing it's hard to give advice on the settings to try here.
Hi there,
Thanks for this open-source OGC-3Dtiles pnts tiler, pretty fast and customizable CLI! Also pretty great that the 2.0.1 version stores coordinates and transforms in local CRS that is then transformed to EPSG:4978 via only the root transform.
We produce some pretty massive tilesets, a few billion points, of photogrammetric 3d-scans done via ground + drone captures - not purely airborne lidar - and I noticed the tiles have a pretty small extent with
maximumScreenSpaceError: 1px
in CesiumJS and there are lots of them. Plus Cesium using a depth-first rather than breadth-first approach at loading children tiles makes the problem pretty apparent in our case - just opened an issue here.I tried enforcing a resolution that matches the points spacing at highest resolution, 0.1mm on some focus areas, to see if the resulting tiles would be bigger, but got the error message
gocesiumtiler/cmd/main.go
Lines 202 to 203 in adb1649
resolution should be between 1 and 1000 meters
Would it be possible to allow for resolution of 0.0001m (0.1 mm)? Do you think this could solve the described problem, or maybe I should try enforcing a
min-points-per-tile
of 10k or 20k?The text was updated successfully, but these errors were encountered: