DigiFarm Leverages HPC to Drive the Future of Precision Agriculture in Europe and Beyond

Oslo, June 2025

DigiFarm, a trailblazer in agricultural technology, is transforming the way farmland is understood and managed. By combining satellite imagery with state-of-the-art deep learning models, DigiFarm delivers hyper-accurate field boundary detection at scale. But reaching this level of precision and scalability wasn’t possible without high-performance computing (HPC).

Thanks to support from the EuroHPC Joint Undertaking (JU), EOSC-DIH, and access to some of the world’s most advanced supercomputers, DigiFarm has been able to train and deploy large AI models across millions of hectares of agricultural land in Europe.

The Problem: Inaccurate Field Boundaries are Holding Back Digital Agriculture

Accurate, up-to-date field boundaries are essential for the success of modern precision agriculture. Yet in many parts of the EU and globally, cadastral data such as LPIS (Land Parcel Identification Systems) and CLU (Common Land Units) are often outdated by 10–15 years. This leads to:

• Manual mapping that is time-consuming and error-prone

• Wasted resources and lower crop yields

• Poor traceability in food supply chains

• Missed opportunities for carbon tracking and climate-smart agriculture

DigiFarm set out to change this by using Earth observation data (e.g., Sentinel-2 imagery) and deep learning to generate field boundaries at a resolution of 1 meter—with over 96% accuracy.

The Solution: Harnessing Supercomputing Power to Train Deep Learning Models at Scale

To achieve this, DigiFarm needed access to massive GPU compute power – far beyond what is possible with standard cloud services or in-house infrastructure. By working with the EOSC-DIH and applying to EuroHPC JU’s Benchmark and Development Access call, DigiFarm gained access to several Tier-0 and Tier-1 EuroHPC systems, including:

• LUMI-G (CSC Finland) – one of the world’s fastest GPU partitions

• MeluXina (LuxProvide, Luxembourg)

• Vega (IZUM, Slovenia)

These systems enabled DigiFarm’s machine learning team to:

• Process and analyze satellite data across 92+ million hectares

• Train segmentation models for field boundaries across Germany, Austria, Belgium, and the UK

• Achieve Intersection over Union (IoU) accuracy scores of 94–96%, significantly outperforming public cadastral baselines by 12–15%

• Scale up inference pipelines to generate accurate field boundary predictions across entire countries in days – not months

The Results: Practical Impact for Farmers, AgTech Partners & the Planet

With the power of HPC, DigiFarm was able to deliver concrete results for agriculture stakeholders:

✅ Highly Accurate Field Mapping at Scale

DigiFarm’s deep learning pipeline now supports seamless processing of satellite data to deliver precise field boundaries for millions of hectares.

✅ Faster, More Cost-Effective Deployment

HPC access eliminated the need for internal GPU clusters or expensive cloud compute, allowing DigiFarm to scale without high overhead.

✅ Enabling Climate-Smart Agriculture

Accurate field segmentation is foundational for sustainable farming practices – from carbon tracking to improved yield forecasting and resource optimization.

✅ Boosting Traceability and Transparency

Better field maps support agri-food supply chains, carbon registries, and sustainability audits with verifiable, spatially accurate data.

Why HPC Was Critical for DigiFarm’s Success

Through the EuroHPC initiative, DigiFarm received not just compute time but also expert technical onboarding. The support provided by HPC centers like CSC (Finland) and LuxProvide (Luxembourg) helped DigiFarm optimize model training, fine-tune memory management, and accelerate training times on complex GPU systems.

This allowed the team to build and iterate on AI models faster, reducing training time for one country’s model from weeks to days.

The Road Ahead: Scaling Globally

Building on the success of the HPC-powered model, DigiFarm is now working with industry leaders like Bayer, Corteva, and KWS to bring this innovation to scale.

Future goals include:

• Expanding boundary predictions to all 27 EU countries

• Supporting carbon modeling and MRV platforms for regenerative agriculture

• Bringing field intelligence to smallholder farmers in emerging markets such as Tanzania, Ethiopia, and India

  
  

Final Thoughts: HPC + AI = Smarter Farming

At DigiFarm, we believe that innovation in AI and access to supercomputing resources are essential to tackling the food security, sustainability, and climate challenges of the 21st century.