Hi again. I was wondering if there is a tool for doing the reverse of GDAL2Tiles, meaning once I have predicted on the tiles, I would like to output the segmented mask as a large tiff, same way it was inputted to me. Do you know how may I do that? (I see many images here that look like it has been done)

Thank you, this helped a lot. I am now trying to manage the post-processing stage. I am doing this over buildings (my own dataset) and in zoom 20. Is it still true that the post-processing stages are adjusted only to z=20 and for the parking data? Are the adjustments needed something I could manage to do by working out the (features, merge, dedupe) code or is it not recommended to dive into it at this point?

Update, this comes from the rs method. the “cover” stage i have done manually using an adjusted script to my geojson, not using rs. So rs gives me a Segmentation fault. Any idea why?

Hi Daniel, amazing job and thank you for this article. I am currently trying to apply this code on my own pictures. steps i did is a s follows: 1. tile my geoTiff using GDAL2tiles 2. obtain a geojson file 3. use “rs cover” to create buildings.tiles (this was a mess, had to amend my geojson a lot for that to work) 4. run “rs rasterize”

I am stuck here. I am receiving “Segmentation fault (core dumped)” from the command: ./rs rasterize –dataset config/dataset-building.toml –zoom 20 –size 256 file.geojson buildings.tiles masks