HarvestNet Dataset

Detect smallholder farm presence in data-scarce, conflict-affected, or remote regions - Improve accuracy of cropland mapping beyond traditional land cover products (e.g., ESA CCI, MODIS) - Enable timely food security assessments and early warning systems - Support humanitarian aid targeting and agricultural policy design Notes / challenges: Geographically limited to two Ethiopian regions (Tigray & Amhara) - No temporal series — snapshot from 2020–2023, not annual - No crop-type classification — only binary farm/non-farm - Low-resolution imagery (PlanetScope) may miss very small or obscured fields - Requires geospatial and ML expertise to process and interpret