AI in coffee: Intelligent and awake
Contributing almost a third of foreign export earnings, coffee is one of the main cash crops in Uganda. Nowadays, many small-holder farmers, who rely on coffee farming for their livelihood, are facing challenges such as unpredictable weather and volatile market prices. Traditionally, they have relied on manual yield estimates, but these are time-intensive to perform and not always reliable, leaving many farmers unable to safely plan their harvests or manage their resources effectively.
One of the many people affected by climate change is Herbert Katongole, a smallholder farmer from western Uganda. He is a participant in the “Croppie” project and has been involved in testing a new approach to tackling these challenges.
Croppie has helped me to learn about agronomic practices that I didn’t previously know were so important for my coffee. Thanks to the yield estimation and agronomic tips received, I was able to control pests and diseases more effectively.
Herbert Katongole, a smallholder farmer from western Uganda
“Croppie” and how it works
Supported through the BMZ/GIZ-initiative FAIR Forward, Producers Direct, an NGO run by smallholder farmers, the International Centre for Tropical Agriculture (CIAT), in alliance with Bioversity International and M-Omulimisa developed ‘Crop Yield Predictions for Smallholders’, shortened Croppie, a tool powered by artificial intelligence (AI) that helps coffee farmers to predict their coffee yield.
The tool harnesses AI technology to give farmers real-time insights, offering personalised yield predictions and agronomic advice. That way Croppie helps farmers to tackle climate challenges and improve their practices.
Farmers snap photos of their coffee plants and Croppie’s AI-supported image recognition counts the coffee cherries and estimates the likely yield for each plot. This turns simple photos into valuable data, allowing farmers to plan harvests more accurately, manage their finances better, and schedule their activities for maximum profitability.
From 2020-2022, a group of 2,392 farmers in Colombia and Peru have contributed to the dataset for the first time. Starting in 2022, the GIZ FAIR Forward initiative collaborated with the Croppie partners to expand to Uganda. For this purpose, Producers Direct and CIAT have teamed up with the Ugandan agricultural technology company M-Omulimisa. 359 farmers collected over 1,121 pictures and thereby helped to create a representative dataset for the Ugandan coffee sector. This massive data collection has trained Croppie’s AI to make accurate predictions for the local context. That way farmers benefit from AI-based yield predictions and personalised agronomic advice delivered via SMS.
KI meets local knowledge
By combining cutting-edge technology with local knowledge, Croppie is bridging the gap between data and real-world decision-making, helping farmers like James Namara from Western Uganda.
Croppie’s advice helps me remember exactly what to do on my farm at the right time.
James Namara, a farmer from Western Uganda
To ensure that Croppie has a long-term impact, the key building blocks like the data set and model are published as open source on Hugging Face. This makes Croppie a digital public good and ensures that the local ecosystem can access it.
To make the tool easier to use, Croppie is testing a hybrid model that combines digital tools with hands-on support from local promoter farmers, ensuring that farmers can easily apply the advice they receive. Beyond agronomy, Croppie is also exploring how AI-generated data can help farmers access financial services, such as loans and insurance, using the yield data calculated by Croppie as proof of future productivity.
What’s next
Through partnerships with producer organisations in Peru, Colombia, and Uganda, Croppie has gathered local data and insights on what features provide most value to farmers. Some have noted that while the advice is helpful, they’re eager for more tailored insights specific to their unique circumstances. As Croppie evolves, the goal is to provide highly customised recommendations that match individual farm profiles and add new features like USSD and voice call functionalities so that farmers will have greater control over the information they receive, helping them make better decisions on their farms. This powerful blend of AI and farmer expertise is opening up exciting new possibilities for smallholder farmers, unlocking their potential and boosting productivity across the region.
