Idea
A satellite data and deep learning platform mapping smallholder crop fields nationwide to support agricultural planning and policy makers
Research Paper
Core Innovation
This paper introduces a scalable method combining very high resolution satellite imagery with deep transfer learning to accurately delineate 21 million smallholder crop fields at national scale in Mozambique. Unlike prior approaches limited to coarse or regional data, this method achieves 93% accuracy and reveals detailed spatial patterns of field size variation, enabling new socio-economic and environmental insights.
Market Size (TAM)
$2–10B TAM, $1–2B SAM; assumption: global agricultural monitoring and precision farming markets expanding with satellite data adoption
Potential Customers & Pain Points
- Agricultural Ministries needing accurate crop field data
- NGOs supporting smallholder farmers lacking detailed land use maps
- Agribusinesses requiring precise field boundaries for supply chain management
- Environmental agencies monitoring land use change
- Researchers studying agricultural patterns in developing countries
Business Model
Subscription-based API and data platform offering crop field boundary datasets and analytics to governments, NGOs, and agribusinesses
Competitive Landscape
- Descartes Labs
- Planet Labs
- SatSure
Implementation Challenges
- Access to consistent high-resolution satellite data globally
- Adapting models to diverse agricultural landscapes
- Integration with existing agricultural data systems
Validation Strategy
- Pilot deployment with Mozambique agricultural ministry
- Field validation with local agronomists and NGOs
- Performance benchmarking against existing land use datasets
Research Paper Overview
National level satellite-based crop field inventories in smallholder landscapes
Summary
This paper presents a method integrating very high spatial resolution Earth observation data and deep transfer learning to delineate crop fields at national scale in Mozambique, producing a dataset of 21 million fields with 93% accuracy. It reveals detailed spatial distribution and size of smallholder fields, highlighting socio-economic and environmental implications of field size variation across diverse farming systems.