Skip to main content

I'm exploring the new Geopackage tiles writer in FME2017. I would like to to create a GeoPackage with the "Use incoming feature" tiling option. In this case the projection should be UTM35. This works fine, but the option to generate additional zoom levels is disabled. This means that I have to use gdaladdo.exe on the GeoPackage file afterwards to generate additional zoom levels (overviews).

Is there any way to create a GeoPackage with custom tiling and additional zoom levels?

We didn't include additional zoom level generation for custom tiling schemes in the first iteration of the writer because it introduces some extra complexity. Generating these zoom levels begs questions like:

What do we use as a base zoom level?

  • A resampled raster where one of the image dimensions is 1 pixel?
  • A resampled raster that fits in a single 256x256 tile?
  • A resampled raster at a particular resolution? (possibly similar to other tiling schemes?)

Your use case and your input would be valuable in helping us determine what kind of options to expose.


We didn't include additional zoom level generation for custom tiling schemes in the first iteration of the writer because it introduces some extra complexity. Generating these zoom levels begs questions like:

What do we use as a base zoom level?

  • A resampled raster where one of the image dimensions is 1 pixel?
  • A resampled raster that fits in a single 256x256 tile?
  • A resampled raster at a particular resolution? (possibly similar to other tiling schemes?)

Your use case and your input would be valuable in helping us determine what kind of options to expose.

Thanks for the reply! We have implemented this by dividing the dataset into tiles of say 256 by 256 and taking the nearest power of two that is higher than the number of tiles as max level. For example if the image would result in 80 x 100 tiles, we'd say that this is level 7 (128 x 128 tiles) and generate overviews up to level 6. All empty tiles will of course be dropped from the tile set.

 

 


Thanks for the reply! We have implemented this by dividing the dataset into tiles of say 256 by 256 and taking the nearest power of two that is higher than the number of tiles as max level. For example if the image would result in 80 x 100 tiles, we'd say that this is level 7 (128 x 128 tiles) and generate overviews up to level 6. All empty tiles will of course be dropped from the tile set.

 

 

That sounds pretty reasonable to me, and I think it's equivalent to the second option I mentioned in the answer (a resampling that fits in a single tile).

 

I imagine we'll also include an option to tie the zoom levels to particular resolutions, but your method would probably make a good default.

 

 


Reply