The Artificial Intelligence benchmark datasets for Land Cover Classification from Satellite Imagery (AI4LCC) dataset collection is an initiative of the Continental Surfaces Data and Services Hub –  THEIA, part of the DATA TERRA Research Infrastructure for the distribution of AI-based training sets for the classification of landcover from satellite imagery. The datasets can be used to train a classical machine learning or more advanced Deep learning algorithm to process information.
Currently, the collection consists of the following datasets:

1) Collection MultiSenGE for multi-temporal and multi-modal landcover classification, with 8157 multi-temporal patches of Sentinel-1 and Sentinel-2 imagery (256x256) over the Grand-Est region (France). The collection is organized in a standard procedure described in the metadata. The products are available at :
- The metadata : AI4LCC-MultiSenGE.json
- The Sentinel-1 temporal serie patches (GRD) : Sentinel-1 patches
- The Sentinel-2 temporal serie patches (L2A) : Sentinel-2 patches
- Ground reference patches : Ground reference patches
- JSON files for each patch : label files

2) Collection MultiSenNA for multi-temporal and multi-modal landcover classification, with 12258 multi-temporal patches of Sentinel-1 and Sentinel-2 imagery (256x256) over the Nouvelle-Aquitaine region (France). The collection is organized in a standard procedure described in the metadata. The products are available at :
- The metadata : AI4LCC-MultiSenNA.json
- The Sentinel-1 temporal serie patches (GRD) : Sentinel-1 patches
- The Sentinel-2 temporal serie patches (L2A) : Sentinel-2 patches
- Ground reference patches : Ground reference patches
- JSON files for each patch : label files

Information on these collections is available in Wenger & al., 2022^[1] and Wenger & al., 2022^[2]
The PhD thesis related to this work is available here Wenger, 2023^[3]
A scientific publication related to this work is available here Wenger, 2025^[4]
In addition, usefull Python tools can be found on Github to extract information on the dataset.

The AI4LCC dataset products are under the open source license Creative Commons License - Attribution Non Commercial 4.0 International (CC-BY-NC 4.0) Creative Commons License - Attribution Non Commercial 4.0 International .

The collection is disseminated using FAIR principles by the DATA TERRA Research Infrastructure (Continental Surfaces Data Hub – THEIA) through the diffusion service EOST/A2S hosted at University of Strasbourg.