Obtaining raw training data


I’m a computer vision researcher and looking to build some datasets for various crop types across the world, in order to measure generalization of different models. I understand that this L4A processor works with a pre-existing RF algorithm, but I’d like to try a wide range of different models.

My question is: is there anywhere I can obtain the raw shapefiles with the individual parcels labeled by crop type? I’ve searched all over for LPIS data, and found small handfuls on various countries’ government websites, but given that this initiative has already put in a lot of the effort to combine this information, I’d love to be able to download it to use for my own research.

Thank you,

Hello Nick.

The data that you are refering to are owned by the member states of the European Union. They are collected in the framework of the European Common Agricultural Policy (CAP). You can find some of these data on the national/regional geoportals of these member states. Sen4CAP is a system that was developed to show the usefulness of the use of Sentinel data in the CAP monitoring context. The L4A processor doesn’t use any pre-existing RF algorithm; it uses part of the input parcels to train the RF algorithm.


Thank you for the clarification Philippe.

I understand the CAP framework and individual countries’ implementation of LPIS (such as SIGPAC in Spain), but I’ve had trouble finding many crop type parcel shapefile datasets from national/regional geoportals so was hoping this program had some central repository. If not, I’ll continue searching for these datasets but would appreciate any suggestions for specific repositories if possible. I’ve come across a handful on https://www.europeandataportal.eu/en if anyone else is curious.

I also understand that the user provides the data for the L4A to train a RF model with, but I’m looking at using different algorithms with different features.