Foundation Models
Foundation Models
The Three Foundation Models (AgriFMs), complementary AI systems spanning environment, management, and plant growth
The Three Foundation Models (AgriFMs)
Three complementary foundation models, one shared framework
AgriScienceFM develops three foundation models, each learning from a distinct family of agricultural data, and together forming a reusable basis for a wide range of downstream scientific and real-world applications
Environment Foundation Model
AgriFM-E
A spatio-temporal AI model trained on satellite imagery, weather, soil, and crop phenology data
AgriFM-E focuses on the environmental drivers of agriculture. It will learn from Earth observation, weather, soil, crop calendars and other spatio-temporal data to better understand how environmental conditions shape plant growth, crop development and agricultural land use. This model will support downstream tasks such as crop type mapping, crop and water monitoring, yield outlooks, soil-related insights and farm advisory applications
APPLICATIONS INCLUDE
Management Foundation Model
AgriFM-M
An agricultural document-understanding model trained on agronomy literature, advisory services, extension materials, and agricultural knowledge
AgriFM-M focuses on agricultural management knowledge and decision-making. It will learn from advisory materials, scientific literature, agronomy textbooks, policy documents and other agricultural knowledge sources to extract, interpret and reason over management-related information. This model will support downstream tasks such as soil health advice, nutrient and irrigation guidance, farm advisory support and evidence-based agronomic decision-making
APPLICATIONS INCLUDE
Genetics & Growth Foundation Model
AgriFM-G
A plant imaging AI model trained on multimodal plant phenotyping and imaging datasets
AgriFM-G focuses on plant structure, growth and phenotyping. It will learn from plant image libraries, UAV imagery, field sensor data, RGB, depth, 3D and multispectral images, together with knowledge about plant development and crop growth. This model will support downstream tasks such as plant disease detection, crop health monitoring, high-throughput phenotyping, stress detection and breeding for climate-resilient crops
APPLICATIONS INCLUDE
See the models in action
Explore how AgriFM-E, -M and -G power the project's four scientific use cases and its open deliverables
Committed to open science, interdisciplinary collaboration, and responsible AI innovation for sustainable agriculture
FUNDING
Funded under the Horizon Europe programme
© 2026 AgriScienceFM · Privacy & Cookies
Open · Trustworthy · Explainable