Tile AGIS imagery

gistfile1.txt

The original imagery has 4 bands (RGB and NIR) and 16bit values

Build the virtual image using the RGB bands

gdalbuildvrt -a_srs EPSG:21781 -allow_projection_difference -b 1 -b 2 -b 3 -input_file_list tifs16.list 2016rgb.vrt

Translate from 16 bit to 8 bit values (note this is fairly slow)

gdal_translate -scale -of VRT -ot Byte 2016rgb.vrt 2016rgb8.vrt

https://github.com/GitHubRGI/geopackage-python has the parallel gdal2tiles version

gdal2tiles_parallel.py -s EPSG:21781 -z 8-19 -p mercator -f PNG --processes 4 2016rgb8.vrt

... wait a long time (on a 6 core xeon machine using --processes 6 ~47h down to zoom 20 which however doesn't make sense for 25cm/pixel imagery, which takes 28 hours to zoom 19) ... 

generate mbtiles file:

set SQLITE_TMPDIR to a location that has at least enough space for a full copy of the mbtiles file (vacuum needs this)

https://github.com/mapbox/mbutil

mb-util --image_format=png --scheme=tms 2016rgb8/ /server/agis_2016.mbtiles

Takes roughly 20 hours (but slow disks).