search

fieldsoftheworld / ftw-baselines

Code for running baseline models/experiments with the Fields of The World dataset
https://fieldsofthe.world/
MIT License
58 stars 4 forks source link

include time information in the input raster format for inference #26

Closed cholmes closed 3 weeks ago

cholmes commented 1 month ago

Using ftw inference download (when #25 merges) results in a file that loses all time information. If we want to do proper fiboa output and include determination_datetime (which we should, as it's a good best practice), we should use the time stamp of the latter satellite image. So we need to include that in the input file format in some way. It is in the STAC Item, so we just need to grab it and stick it somewhere that we can then include in the inference output and polygonization.

m-mohr commented 3 weeks ago

@cholmes Why is it the datetime of the latter image? Is it always window B? Or actually the later datetime (assuming window a and b have no particular order)?

cholmes commented 3 weeks ago

@cholmes Why is it the datetime of the latter image? Is it always window B? Or actually the later datetime (assuming window a and b have no particular order)?

Hrm, I'm not actually sure if it's a requirement that window a and b enforce an order / make an error. We should probably check that case and either handle it or error out. But yeah, it should be the latter datetime (based on the fact determination_datetime is the last known time that you deem the boundary as accurate)

m-mohr commented 3 weeks ago

Oh, that makes sense. I can just grab the latter date from STAC then.

There's currently no check on window A before window B temporally, I think.