CatCrops_identification: A Python Project for Early Crop Type Classification Using Remote Sensing and Ancillary Data

CatCrops_identification is a Python library developed for the early classification of crop types using remote sensing data (Sentinel-2) and ancillary information. It is based on a Transformer model adapted for the analysis of spectral time series with variable length, and it allows the integration of auxiliary data such as the previous year’s crop, irrigation system, cloud cover, elevation, and other geographic features.

The library provides tools to download and prepare datasets, train deep learning models, and generate vector maps with plot-level classification. CatCrops_identification includes scripts to automate the entire workflow and offers a public dataset that combines declared and inspected information on crop types in the Lleida region.

This approach improves classification accuracy in the early stages of the agricultural season, offering a robust and efficient tool for agricultural planning and water resource management.