Closed ircwaves closed 6 years ago
Earlier estimate of timing was off. Current ballpark for coregistration is over an hour.
Need to determine where/why it is taking so long. I suggest adding verbose_out
s that include a time delta from the start of the process function.
Probably better to use sentinel2Data.inventory(...)
here, because this fetches the file but doesn't update the database inventory:
https://github.com/Applied-GeoSolutions/gips/blob/5f0dcf9d69ac16591a0ba95049dedeedd38b722f/gips/data/landsat/landsat.py#L1406-L1411
12 minutes of converting jp2 to tif:
1 hour 10 minutes for CookieCutter:
(side note: we need to replace lots of stuff in the sentinel2 driver with VRTs)
For the CookieCutter, set crop=True
in the call. That should make the resulting file significantly smaller.
Instead of using CookieCutter at all, all relevant are gdal_merge
d together, which only takes a few minutes, reducing the processing of a coreg product to about 4 minutes.
And in the event that you've coregistered a product for a given tile+date previously, additional products require very minimal (seconds) of additional processing time.
Add a product option to coregister landsat onto Sentinel-2