Artificial intelligence (AI) is being tested in Uganda to help farmers detect crop diseases at an early stage. It also gives them an instant recommendation on actions needed. According to researcher Owomugisha Godliver, AI technology could improve agriculture for smallholder farmers and alleviate food insecurity in Africa.

About 35-40 percent of all agricultural yields in sub-Saharan Africa are lost annually because of crop pests and diseases. Much of this loss is because it takes too long for smallholder farmers to get answers from agricultural experts about what is wrong and how to treat affected crops.

At Busitema University in Uganda’s Eastern region, a research project aims to tackle this problem by using AI software installed on low-cost smartphones. Around 100 cassava smallholder farmers in the region have participated in the study.

“The farmer takes a picture of the crop with her phone and the AI-based application compares the photo with 2,756 pictures of three cassava diseases and two types of pest damage. Thus, the farmer will know directly if there is something wrong with the crop”, computer scientist Godliver, who leads the research team, explains.

The lack of an available database of agronomic questions and answers about staple food crops in sub-Saharan Africa made it more complicated to develop instant recommendations. The researchers interviewed all the farmers in the study to find out what questions they had about their crops and farming techniques. Agricultural experts then provided answers to the questions, which the researchers compiled in a single dataset of nearly 4,000 question-answer pairs.

“The real-time feedback on their smartphones is an important tool for smallholder farmers. Now they can diagnose without having to wait for an expert to make a field visit, by which time the harvest could already be destroyed”, Godliver observes.

The research team is currently working on a version of the AI-based mobile app in the local Luganda language. This will make the system more accessible and empower smallholder farmers to make informed decisions for better agricultural outcomes, Godliver points out. In addition, an important feature is that the app works offline, so farmers living in remote areas with limited internet connectivity can use it.

The researchers have also been looking at ways to detect crop diseases before there are any visual symptoms. This can be done by using a spectrometer that spots changes in light absorption on plants. “By using this tool, farmers can ensure they are planting healthy seedlings, which reduces the risk of late-stage disease detection”, Godliver says.

Normally, a spectrometer costs around 1,000 US dollars, which puts it out of reach for most smallholder farmers. However, the researchers have developed a portable smartphone spectrometer at a low cost of 5-8 US dollars.

“Our research approach has from the beginning been that the AI models we develop work in low-resource environments”, Godliver concludes.

AI technologies present several transformative opportunities for African smallholder farmers, according to NAI researcher and food systems expert Assem Abu Hatab. "Decision support tools provide essential insights into crop selection, pricing, and market timing, while digital platforms enable access to financial services, crop insurance, and market information. These precision-driven tools lower costs, improve food security, and lessen environmental impact", Abu Hatab says.

Education in Africa can advance with the help of AI

AI can also be an important tool for addressing challenges in education systems on the African continent. One problem is that many children face difficulties learning in school because lessons are in English, French or Portuguese rather than the local language they speak at home.

Oladipupo Sennaike is a computer scientist at the University of Lagos in Nigeria and the principal investigator for the EduAI hub (Hub for Responsible Artificial Intelligence for Education Research Network in Africa), which runs several research projects on the continent.

“One of our projects aims to develop an AI-driven solution to support the teaching of mathematics in Nigerian primary schools. I know this will be helpful, because a colleague once went to teach maths in the north. He didn’t speak the local language, Hausa, and the students didn’t know English”, Sennaike remarks.

Other EduAI hub research projects include an improved learning environment for disabled students in higher education, developing a sign-to-text and text-to-sign system that will help communication between students with hearing impairments and their teachers, and a text-to-speech system for visually impaired and blind students.

“Everybody in the research teams is excited by how much we can achieve, and in the schools both students and teachers are embracing the AI technology”, Sennaike states.

But there are also concerns, he adds. One regards infrastructure, and computing power, which needs to be better and faster — which also means costlier. Nevertheless, it is necessary “because you can do only so much, if you don’t have access to such computing power”.

Another concern is about existing data on African contexts. If there is not enough data, the AI models will not be good enough.

“Does what is coming out of the computer represent Africa and is it a cultural fit for here? Because without data from our environment, we will be relying on solutions imported from elsewhere that are definitely not going to work for us”, Sennaike concludes.

TEXT: Johan Sävström