Building a Model to Rapidly Identify Disease in Ghana’s Maize Plants

Disclaimer: Aya Data holds to the confidentiality agreements made with our clients. We shall not disclose any confidential information regarding clients and/or projects, pursuant to the individual agreements made. Consequently, client and project names and any identifiable information may be kept anonymous in our Case Studies.

Overview

Maize is one of the most important crops for Ghana’s agricultural industry; there have been significant, largely successful, investments into finding solutions to increase yields over the years. However, maize diseases still pose a significant challenge for maize farmers and communities.

Aya Data collaborated with Demeter Ghana on a pro-bono initiative aimed at addressing this challenge. The purpose was to develop a computer vision model which could accurately detect diseases in maize plant leaves.

Data Annotation, 

Data Science and AI

Image segmentation

Industry

AgTech

Headquarters

London, UK

Company Size

250+

Challenge

Common maize diseases were often misdiagnosed and mistreated in rural areas of Ghana, which Demeter Group sought to address. One of the main issues is that many diseases that affect the maize plant resemble each other, which makes it challenging for farmers with limited agricultural training to identify them correctly.

Agronomists are rare and expensive in these regions, so relying on them for advice and support on a farm-to-farm basis is not a sustainable solution. The challenge was to create an accessible model that could accurately diagnose common diseases through an app and provide recommendations for treatment.

Contact us

Solution

Aya Data sought to label 5,000 images captured with smartphones with the help of agronomic experts in Ghana. The dataset contained images of both diseased and healthy maize plants. The diseases were bounded and tagged with specific classes, which indicated the disease type and severity.

Using this labeled dataset, Aya Data trained an Artificial Neural Network (ANN) and deployed it within a basic smartphone application.

Results

The results of the collaboration between Aya Data and Demeter Ghana were:

  1. The development of the diagnostic app: Demeter Ghana developed and successfully distributed the app to its cusomters, allowing farmers to submit images of maize plants for disease detection.
  2. High-accuracy detection: The developed model had an impressive 96% accuracy in identifying disease types and severity levels in maize plants from the images.

By providing farmers with a tool to accurately identify and diagnose diseases, the app enabled timely intervention and appropriate treatment, mitigating the impact of crop diseases on maize yields.

This pro-bono initiative supported Demeter Ghana in their mission to empower farmers and enhance agricultural practices by leveraging computer vision technology.