Teaching Smartphones How to Predict Plant Disease

In our first-ever solicitation for grant proposals, we’re very fortunate to have received hundreds of applications detailing fascinating and worthy research and development projects for improving smallholder farmers’ lives. We interviewed one of the applicants for this story.

Neema Mduma, Ph.D., is a lecturer in the School of Computational and Communication Sciences and Engineering at the Nelson Mandela African Institution of Science and Technology in Arusha, Tanzania. She also runs an organization promoting science and technology for girls in primary and secondary schools.

Mduma is an expert on artificial intelligence and machine learning. Inspired by the success of her Ph.D. research project, Mduma is now planning to build an elegant yet easy-to-use machine-learning tool to capture early signs of pests and diseases in crops, allowing farmers to respond in a low-cost and environmentally friendly manner.

The concept

Mduma explained to us in an interview that she’s designing a smartphone application for initial deployment on farms in Tanzania and Uganda. The app would allow farmers to detect and predict threats to their crops by taking pictures of plants that look infected. Building the app will require an enormous amount of computing time processing thousands of images of healthy and infected crops alike. Mduma says she then has to then teach an artificial intelligence program how to notice crop diseases and predict just how bad a noticed infection might get or how far it may spread. The tool would also offer farmers advice on how best to treat problems or head off threats, all by just scanning and processing the images farmers capture in the field.

During our conversation with her, Mduma kindly and patiently explained the inspiration for her research and how the technology she is designing would work in theory and practice.

Dr. Neema Mduma speaks with Grow Further

Q: Where did you come up with the idea for this research proposal?

A: My Ph.D. research was focused on developing machine learning tools for predicting student drop-outs, in order to predict whether a student is going to drop or not and what can be done to address, to rescue that situation.

Then after that, I worked on different projects, particularly on artificial intelligence in agriculture.

Q: Does the technology for predicting student drop-outs work the same for predicting crop diseases?

A: Oh yes, because when you are talking about artificial intelligence, inside artificial intelligence there is machine learning, and for machine learning, there are different functionalities which can be done. In my Ph.D. studies my issue was prediction, focusing on using machine learning and using algorithms focused on prediction.

The technology is still there; the core function is still there.

Q: Could you tell us a little bit more about the research project you and your team proposed in your application to Grow Further?

A: “We are working on detecting diseases in maize and common beans. As you know, there are so many diseases, but we are focused on the diseases that are mainly affecting productivity in Tanzania and other parts of sub-Saharan Africa. So for maize, our target is maize lethal necrosis and streak virus, and for beans, our focus is on bean rust and bean anthracnose. These are the diseases that are highly affecting productivity, especially for smallholder farmers.

Our target here is to develop a tool that can be deployed in a mobile application.

For maize, our target is maize lethal necrosis and streak virus, and for beans, our focus is on bean anthracnose. These are the diseases that are highly affecting productivity, especially for smallholder famers.

A major undertaking

Q: What’s the first step toward developing this mobile phone app?

A: We ran into one challenge, in fact, one of the biggest challenges facing machine learning developers, and that is a lack of datasets.

Our focus first is to collect images, thousands of images, of both healthy and disease-affected plants. And then our plan is to put the data in an open-access registry, so it’s not going to only benefit us, but other machine learning researchers can use our data to develop their datasets to come up with different solutions, maybe from different countries.

Q: With this app, a farmer would just take a picture of plants and the tool would warn of any signs of disease?

A: Yes, that is what we call detection, early detection of disease.

We have already developed machine learning tools to help farmers in different sectors of agriculture, detecting other diseases affecting other crops like banana and Irish potato, and there we have seen promising results.

Q: The data you’ll be processing will be made openly available to everyone?

A: Absolutely. Even other researchers all over the world who are applying machine learning techniques, they’ll be able to maybe come up with improved solutions or even try using the datasets they have toward other challenges, other [plant] diseases in their country.

Challenges ahead

Q: What do you think will be the most difficult part of this R&D effort?

A: I think maybe the most difficult part could be how we’re going to deal with errors.

First of all, it will require a long period of time. It could be maybe three to four months, or five, to collect and analyze all those datasets from different parts of Tanzania. And then there’s also the issue of geocodes, because sometimes when you go to collect data in rural areas the internet will not be there or won’t be as accurate, so we’re planning to develop a way to ensure GPS precision using other techniques, maybe even triangular location.

We’ve already planned for these challenges because we faced the same challenges in the collection of data for the banana and Irish potato projects.

Q: Will enough smallholder farmers have the right kind of phones to take advantage of your innovation?

A: We do understand the issue of mobile phone penetration in Africa and Tanzania. So far, around 70% of people have mobile phones. But sometimes in certain areas, they have maybe a mobile phone but maybe not a smartphone…the answer is to provide mobile applications on smartphones for use by their village leaders or even the farmers’ groups.

Q: How do you plan to scale up this innovation?

A: We’re planning to have partnerships with different organizations such as agricultural research institutes, and we’re working with the Tanzania Agricultural Research Institute. And we’ll also work with farmers’ communities, different sectors, NGOs, and the Ministry of Agriculture.

Grow Further

Photo credit: Dr. Neema Mduma collecting data in the field. Nelson Mandela African Institution of Science and Technology/Neema Mduma.

Subscribe to Our Newsletter