Project title
Machine Learning Tools for Early Detection of Maize and Common Bean Diseases for Climate Change Adaptation
Grantee institution
The Nelson Mandela African Institution of Science and Technology (NM-AIST), a graduate university in Tanzania.
Team
Neema Mduma, Lecturer in Information and Communication Sciences and Engineering, NM-AIST.
Angela Mkindi, Lecturer in Life Sciences and Bio-engineering, NM-AIST.
Ritha Ituwe, Accountant, NM-AIST.
Key Partners
Mbeya University of Science and Technology, data collection.
Tanzania Agricultural Research Institute, agricultural expertise.
Rift Valley Cooperative Union, farmer’s group.
Makerere University, machine learning.
Why We Exist
Background
In Tanzania, the agricultural sector, largely dominated by smallholder farming, accounts for one quarter of the national GDP, but production is stagnant, while the population is expected to double by 2050. The effects of climate change are deepening the vulnerability of agriculture to disasters. Chronic malnutrition rates are above the African average, with 32% of children under 5 being stunted.
Maize is the most important staple crop in Tanzania, used to make ugali among other dishes. Common beans also play an important role in nutrition and food security. Both crops see significant losses to pests and diseases, up to 100% on some smallholder farms. These problems can be managed, but early identification and intervention is important. Very few farmers in Tanzania are using software to identify diseases of maize or beans. However, 67,000 farmers are now using a banana disease identification app developed by the applicant team.
Approach
The project will assess needs, train and test a machine learning system for identifying diseases of maize and common beans and develop a mobile application. The mobile application will be distributed through farmer associations to reach farmers who do not own smartphones.
Targets
Current status
NM-AIST has developed a similar app for banana diseases that farmers have adopted, but existing solutions for maize and common beans are not being used by farmers. Existing solutions are trained on data from other parts of the world, lack Swahili support, and in one case require farmers to purchase additional hardware. NM-AIST has identified existing datasets and made plans to collect additional data as part of the project.
Role of Grow Further
Grow Further will provide a grant of approximately $63,000 over 2 years to cover all research and development costs associated with the project and some agricultural extension activities. Larger-scale agricultural extension activities will be covered by government and NGO partners. In the event that the results of the project are widely adopted, Grow Further will arrange a rigorous independent evaluation to determine whether it successfully improved nutrition, farm income, and other socioeconomic variables.
Our long-term plan is to form a network of chapters or imitators so that farmers and scientists have lots of choices when it comes to testing small or unconventional ideas, and so that people who don’t work at a grant-making agency have a meaningful way to get involved.
