Did you know that the Czech Republic is the world’s fourth-largest hops producer, not to mention the global leader in beer consumption per capita—by a significant margin? To the enjoyment of beer lovers everywhere, 2021 went on record as the year of the country’s largest hop harvest in 25 years.

The successful harvest happened despite dramatic fluctuations in crop yields due to climate change and an explosion of pathogens. Today, the top priorities for the agricultural sector are replicating that success and taking steps to guard against future harvest threats, such as droughts and disease.

Agricultural technology, or agtech, may offer the best path forward. Solutions like computer vision, agricultural artificial intelligence, and internet-connected crop monitoring systems are slowly but surely driving a radical transformation of the sector and are poised to change the world of farming—but not without vast amounts of labeled data to train models and deliver insights that empower precision agriculture and smart farming.

In the Czech hops industry, such innovations help farmers address relentless climate change issues and increase crop yields by using data—data for learning from past successes and failures and enhancing future resilience.

To make these wins possible, agtech consulting firm Agritecture teamed up with Microsoft and Asahi, owner of the world-renowned Czech Pilsner Urquell brand. Recently, CloudFactory's Chief Data Science Advisor Keith McCormick sat down with Agritecture’s Henry Gordon-Smith and David Ceaser to talk more about the project. What follows is a summarization of the key points from that discussion.

What challenges are facing the hop sector?

More so than many other crops, hop plants depend on sufficient and consistent moisture levels. With enough water, hop plants can grow 23 feet (seven meters) in two months, but without suitable climate conditions, they languish. In several worst-case scenarios, hop farmers have abandoned entire areas due to reduced groundwater.

Pathogens like the downy mildew parasite also threaten hop-growing regions in the Czech Republic. Although fungicides can counter the effects of downy mildew, farmers can use them only on a limited scale due to EU regulatory compliance, and only when local climate conditions allow.

The unpredictable nature of climate change and the elevated threat of pests and disease that comes with it also leave hop growers especially vulnerable to volatility. That volatility has translated into 20% year-on-year fluctuations in crop yields in the Czech Republic.

But with eyes on advancements made possible by technology, the future for Czech hop farmers looks bright.

One recent development [for hop farmers] is algorithms for environmental conditions. The algorithms can signal that a downy mildew outbreak is coming, which allows the farmer to spray pesticides before the outbreak occurs. And it's a double win because [spraying early] means farmers will use less pesticides than they would in a full outbreak.
- David Ceaser, Lead Agronomist of Agritecture

How does the agricultural sector use annotated data?

Microsoft's AI for Earth initiative aims to apply artificial intelligence and machine learning to improve the way farmers monitor, model, and manage their operations. The initiative has proven a key enabler of precision agriculture and smart farming.

The current focus of the project is to deliver a steady stream of valuable data collected by Internet of Things (IoT) technologies and specialized agtech solutions, such as sap flow sensors and plant stem diameter sensors that accurately monitor plant health in real-time.

The project is about applying AI to hop plants to optimize water supplies and all other conditions. To do this, we've created a galvanizing brand. Under the leadership of Asahi, and after multiple site visits, we chose six different pilot farms to work with. We also narrowed down a selection of equipment—IoT technologies—to produce streams of data. And partners help with some of the algorithms and machine learning models we're building.
- Henry Gordon-Smith, Founder and CEO of Agritecture

And although the Agritecture team is already capturing computer vision data with cameras, they realize that they’ll also need to manage large quantities of manual annotation to fully capitalize on it. Doing so takes three unique skill sets:

  • Subject matter expertise like that held by Agritecture team members, who are physically in the fields, diagnosing the problems, and capturing data.
  • Data annotation expertise from a managed workforce provider like CloudFactory; our data analysts transform the data so computer models can consume it.
  • AI expertise from internal staff or consultants, who rely on the annotated data and use it to build computer models.

Data, both historical and real-time, powers informed decision-making in any industry, including agriculture. Historical data gives vital insights into past successes and failures in terms of crop yields and quality. Real-time data helps farmers detect the onset of pathogens and toggle water levels far sooner than manual inspections, giving them a chance to proactively protect and enhance yields.

How do farmers turn agricultural data into actionable insights?

With cameras monitoring the growth and maturity of crops, farmers are often collecting enormous amounts of unstructured data. In its raw form, that data is of little use; there’s too much of it to make sense of, and you can’t use it to derive machine learning insights until annotations are added.

With annotations comes the potential to use that data for machine learning, which is where the prediction—the real magic—happens. For instance, a hops yield model can help hop growers forecast yields and alter their plans ahead of time. 

Agritecture isn’t there yet. But it is on the way. The day is coming soon when farmers everywhere will collect and process enough computer vision data to be able to use AI to detect problems and protect crops. Yes, it will require manual data labeling on a grand scale. But the benefits will far outweigh the costs. Yield forecasting will help farmers allocate labor requirements. An understanding of trouble indicators will lead to repeatable intervention strategies. And by being empowered to act sooner, farmers will prevent plant stress before it occurs, leading to healthier plants, heartier yields, and happier farmers.

Watch our full interview with the Agritecture team to discover more about how data and AI are transforming agriculture. We also invite you to learn how CloudFactory helps solve the challenges of precision agriculture and smart farming by scaling data preparation and labeling with a fully managed external workforce.

New Call-to-action

Computer Vision AI & Machine Learning Agriculture Data Annotation Drones

Get the latest updates on CloudFactory by subscribing to our blog