Provisional publication of ai4sf_ml.parquet, a field-boundary Parquet file with 439,013 records. Source, license, processing history, and intended use are TODO.
Field boundaries across smallholder farms in Cambodia and Vietnam, with a fixed train/val/test partition for machine learning.
ai4sf_ml.parquet, 439,013 polygon featuresThe split column carries this partition. Use it as given, so that results stay comparable across models trained on this dataset.
The group column preserves the source tile grouping used when the splits were built. Fields that share a tile stay in the same split, which prevents neighbouring fields from leaking information between training and test.
This dataset is stored in EPSG:32648 rather than EPSG:4326. Reproject before combining it with datasets in geographic coordinates.
Please cite the original authors:
Persello, C., Grift, J., Fan, X., Paris, C., Hansch, R., Koeva, M., & Nelson, A. (2023). AI4SmallFarms: A Dataset for Crop Field Delineation in Southeast Asian Smallholder Farms. IEEE Geoscience and Remote Sensing Letters, 20, 2505705. https://doi.org/10.1109/LGRS.2023.3323095
Other field boundary datasets from this organisation are listed at source.coop/ftw.