orion-ai-lab/s4a

Sen4AgriNet: A Sentinel-2 multi-year, multi-country benchmark dataset for crop classification and segmentation with deep learning

s4a's Issues

Issue While Preparing Own NETCDF Dataset for training

Hi, I am preparing my own netcdf dataset for training I am using sentinel2 data with 4 Bands(B02,B03,B04,B08) and I have labels and parcels.I have some crop fields with patch size of 366x366 which include crop area(Area of interest).I have masks to identify area.
Questions:
1.Will the mask (True False) will go in parcels
2.in labels only the pixels of the crop have crop labels(assigned in utils>config.py) used for training?
3.I want to make netcdf file that the algorithm requires but I am facing challenges.Can you provide me the code which creates netcdf files on timeseries.
If you can provide me the code and answer my questions that will be really helpful

Can't find train/test split for Scenario 1

This is advertised as a benchmark dataset. To be a benchmark, it needs to have a well-defined train/test split, but I can't find the train/test splits you used for Scenario 1 from your paper. Without the split, it is difficult to compare methods.

Dropbox download link doesn't work

Hi,

Unfortunately I am unable to download the data using the Dropbox link. Will you making this available?

Thanks

Is data from 2016 to 2018 also open sorce?

Different from 2016 to 2020 mentioned in the article is that the download link you provided on GitHub only has 2019 and 2020. If the data from 2016 to 2018 is also open access, where can I acquire it?