We are Norway's leading agricultural technology (agri-tech) startup. We have created an intelligent software analysis technology utilizing machine learning solution to help farmers optimize their crop production.
We have developed a state-of-the-art deep neural network model for Field Delineation Model leveraging super high-resolution data at 0.25m. Our model has achieved 40x higher accuracy than existing Sentinel-2 at 10m resolution. We are now super resolving Sentinel-2 from 10m down to 1.25m using our proprietary algorithm, which we are injecting into the field delineation model to increase the accuracy, where we've managed to achieve an intersection over union (IoU) of 0.94.
On top of the Field Detection functionality, we are developing a Crop Detection Model (automatic detection of crop planted on fields in-growth season and seeded acres) and Zoning (variable rate technology) for optimizing variable rates of seeding and fertilizing and yield prediction. We are looking for partners in the South East Asia agricultural market for running pilot programs and licensing our technology in a commercial contract.
We are able to provide a demo of our Field Delineation Model, which was applied in Norway. The demo showcases our technology in a sample area (25km x 25km) as compared to existing Sentinel-2 model market leader and Cadastral mapdata. Do feel free to reach out to us for a connection.
The models available in international regions are:
This technology function provides highly accurate and automatically detected field boundaries and seeded acres, historically back to 2015 and during growth-season, near real time, approx. 1.5 months prior to harvesting. The model data can be provided through an API or regular shape file downloads with regularity.
Crop Detection and Classification
This technology function allows for the automatic detection of crops planted on a particular field which is built on top of yield-data from over 14,000 crop producers over a 5-year period in Norway. The alpha version we released in 2019 achieved a 82-83% accuracy, while our latest model released in fall 2020 reached 90% due to the increased accuracy of field delineation model. We can detect crop types back to 2015 and early in each growth season, approx. 31-37 on Zadoks scale, data includes seeded acres. Model can be scaled to international regions subject to available training data.
This technology functions will require Field Delineation as a prerequisite in order to provide a productive and highly accurate zoning tool. The zoning tool will have ability to classify different zones within a field based on low, medium and high productivity zones which enables variable rate technology. The zoning product is developed through our proprietary algorithm based on 5-year NDVI/EVI data along with national yield-data from over 14,000 producers in Norway over a 12-year period which includes 1.255 million daa's coupled with extensive soil sampling and yield-map data.
- Farm management system (FMS)’s applications using models for automatic boundary detection, bypassing manual drawn boundaries by users.
- Access of model, map, data to their end clients, farmers through API.
- Commodity trading: high accuracy data early in season on seeded acres, crops and health.
- Subsidy payment verification process automation: ability to automate on-field control check for farm subsidy payments.
Digital technology enabling the optimizing of crop farming as what we are offering allows farmers to access capabilities and conduct precision farming. While the agri-tech, digital agriculture space is still nascent and growing, such technologies allows for awareness of agricultural productivity and also to maximize output/efficiency through optimization of existing workflows and processes.
Our cutting-edge technology allows for first adopters or integrators of our technology platform or machine learning capabilities to experiment and experience the potential and value of the future of crop farming.