AlphaCrop fuses satellite imagery, drone footage, and IoT sensor data with a self-supervised Agricultural Foundation Model — giving every farmer AI-powered crop-disease detection and real-time field monitoring.
AlphaCrop brings together cutting-edge machine learning with practical agronomic tools in a single, unified platform.
A self-supervised model pre-trained on multi-modal farm data, enabling few-shot transfer to any new crop or region.
Combines satellite multispectral imagery, drone RGB/thermal, and IoT ground-sensor streams into a single representation.
Farmers draw their field boundary on a live map; the platform instantly delivers health scores, disease alerts, and yield forecasts.
Ask questions about your farm in plain language. The LLM layer translates insights from the foundation model into actionable advice.
50,000 annotated samples covering common crop diseases provide rich supervision signals on top of the self-supervised pre-training.
Continuous satellite revisits and IoT telemetry trigger instant notifications before a disease outbreak spreads across the field.
A large-scale self-supervised model learns rich representations from unlabeled farm data before being fine-tuned on labeled disease annotations.
Inspired by advances in language and vision foundation models, AlphaCrop's backbone is pre-trained with masked-patch prediction and contrastive objectives across all three modalities — no labels required.
Multi-modal, multi-crop data from three complementary sensing modalities — each annotated with disease class and crop-health score by agronomists.
Multispectral tiles (Sentinel-2 / Planet) at 10 m resolution with 13 spectral bands including NDVI, EVI, and SWIR indices.
High-resolution RGB and thermal drone flights at field scale, capturing early-stage disease patterns invisible from space.
In-field soil moisture, temperature, humidity, and nutrient sensors providing continuous ground-truth environmental context.
Disease class labels (blast, blight, rust, wilt, healthy) and crop-health scores assigned by certified agronomists.
Fully annotated multi-modal training samples across 12 crop species and 5 continents
No technical skills required. Farmers simply outline their field on the interactive map and AlphaCrop takes care of the rest.
Click points on the satellite basemap to trace your exact field boundary.
AlphaCrop fetches the latest satellite, drone, and IoT data for your polygon.
The foundation model scores every pixel inside your field for disease risk and health.
Receive a health dashboard, disease alerts, and treatment recommendations.
AlphaCrop's natural-language interface bridges the gap between complex AI outputs and practical farming decisions — in any language.
Supports local languages and dialects so every farmer can access AI insights.
Converts model predictions into actionable treatment and spraying prescriptions.
Queries historical data to identify long-term health trends across seasons.
LLM-composed push notifications explain alerts in simple, non-technical language.
A clear, end-to-end pipeline from multi-modal data collection to farmer-facing recommendations.
Satellite images, drone flights, and IoT sensor readings are continuously ingested, geo-referenced, and time-stamped into the AlphaCrop data lake.
The Foundation Model learns visual-spatial-temporal representations from 50K unlabeled multi-modal samples using masked-patch and contrastive objectives.
Expert-annotated disease and health labels are used to fine-tune the model head, enabling precise classification and health scoring.
Via the web app, a farmer traces their field on a satellite basemap. The polygon triggers on-demand inference for that specific region.
Model outputs are passed to the LLM layer, which composes human-readable insights, treatment plans, and alerts delivered via the chat interface or push notification.
Meet the scientists and researchers driving AlphaCrop at the Remote Sensing Lab, Saint Louis University.
Principal Investigator
Co-Investigator - Professor
Co-Investigator - Research Scientist
Co-Investigator - Professor
PhD Student
For collaboration opportunities, project information, or partnership inquiries, please reach out directly to Dr. Vasit Sagan.
