Sentinel-2 Imagery for Crop Identification
Sentinel-2 Imagery for Crop Identification
As part of the AI4AGRI project, a new dataset is now available for researchers focusing on crop identification using remote sensing and machine learning techniques. The dataset consists of Sentinel-2 MSI images acquired between 2020 and 2024 over an area north of Brașov, Romania. It is designed to support two specific tasks:
- Crop Identification with Temporal Generalization: Training models on data from 2020–2023 and testing on data from 2024.
- Early Crop Identification: Identifying crops during the vegetation season, splitting data into training and testing accordingly.
The dataset includes:
- Sentinel-2 GeoTIFF Images: Stored by year, with a resolution of 800 x 450 pixels and 12 spectral bands at 10 m spatial resolution.
- Ground Truth Masks: RGB and labeled masks of agricultural crops, saved in PNG and MAT formats, along with a legend in PDF format.
- 32×32 Pixel Multi-Spectral Patches: Subsets of the imagery for focused analysis, stored in GeoTIFF and MAT formats.
This dataset offers a resource for testing crop identification models under varying temporal and seasonal conditions. It is structured for straightforward integration into machine learning workflows, supporting both generalization and early identification tasks.
https://zenodo.org/records/14283243
μDACIA5 – A Sentinel 2-based Multispectral Dataset for Agricultural Crop Identification Applications over Brasov area, Romania
https://zenodo.org/records/14283243
Soil roughness estimation data set
https://ai4agri.unitbv.ro/wp-content/uploads/2024/07/AI4AGRI_soil_roughness_dataset_2024.zip
Multi-spectral fractal images
https://ai4agri.unitbv.ro/wp-content/uploads/2023/03/low_mid_high_cmplx_MSI.zip