crstn
commented
4 years ago A few thoughts:
- tiles do not need to be georeferenced, as you assumed correctly – they are indeed only loaded by name/path. Other common image formats used for that purpose such as JPEG or PNG actually don't have a (straightforward) way to be georeferenced at all.
- I would focus on getting an existing solution to work rather than writing your own script – as you mentioned, it is not always trivial to figure out which tiles you actually need, so you might end up rendering a ton of tiles that are never loaded
- I don't get the part about the compiler at the end 🤔
TL;DR: Im stuck between two methods of creating tiff-tiles: A basic script cutting the big tiff in smaller pieces, and gdal2tiles script that have been tinkered with to allow it to produce tiffs. The script produce "good" tiffs (when loaded into qgis) whereas the g2t-tiles are loosing their geo-part. I'm unsure if the g2t-issue is be harmless in an webgis setting, since tiles are loaded based on name
While I can create tiff tile using either the TileSplitter script or a slightly manipulated gdal2tiles-script (now able to work with 32bit data, tiff-files and leaflet friendly tiles) they both have problems. Since I haven't yet build the part for receiving the tiles, I have not been able to see if both of these actually are going to cause trouble.
My issue with the Tilesplitter-tiles is that i'm only cutting the bigger tiff into pieces. I'm not providing any metadata and haven't found an effective way to store them correctly. I lack a good way to figure out which tiles are relevant for each zoom level (For instance at zoom level 7 I need tile 87/54). These fold and file names could of course also be found manually
The Gdal-tiles I would presume are sorted correctly into folders, but the tiff files seem to lose their geo-part in the transition. When I load them into Qgis it says that the crs is undefined and all the tiles are stacked on top of each other. I don't know if this i going to be a problem or not, since a regular mbtile wouldn't have this data neither, but just be loaded in based on its name and folder location. When even the compiler is run it also give the error message. Error 1: Buffer too small. This might just be related to the fact that I only have part of India and not the whole world