App ‘trained’ to spot crop disease, alert farmers
Researchers win grant to further test app for smallholders
App diagnoses deadly cassava diseases in field, sends alerts
Roll-out in Africa needs engagement with farmers, says expert
A team of scientists has received US$100,000 grant to refine a mobile application (app) that uses artificial intelligence to diagnose crop diseases, and aims to help millions of African smallholders.
The CGIAR Research Program on Roots, Tubers and Bananas team won the grant during big data conference in Colombia on 21 September as part of the CGIAR Platform for Big Data in Agriculture Inspire Challenges.
The app, to be used against cassava brown streak disease and the cassava mosaic disease, is expected to be rolled out in 2018.
It accurately diagnoses diseases in the field and will combine mobile phone short message service (SMS) alerts to farmers in rural Africa.
David Hughes, associate professor of entomology and biology at US-based Penn State University, who leads the project together with James Legg, a plant virologist with the International Institute of Tropical Agriculture, Tanzania, say the team needs to continue field-testing and improving its user-friendliness.
The app’s conception was in 2012 but got developed in June-September 2017 through about US$300,000 funding from Penn State University, Hughes told SciDev.Net last month (5 October) in an interview.
The app uses a Google programme called TensorFlow that allows machines to train and learn. “We trained it to recognise plant diseases. What the app does in real-time is to assign a score to a video being captured,” said Hughes. “That score is the probability that the plant in the video shows symptoms of one of five diseases or pests.
“We think the most important value we will create will be through [agricultural] extension workers already helping farmers, and most of whom do already own smartphones. It’s realistic to anticipate that [most] farmers in Sub-Saharan Africa will have smartphones capable of running the app within five to ten years.”
According to Hughes, the project’s expansion is aimed at collecting more images to train the machine to identify more diseases in more crops — such as banana, sweet potato and yam — as well as work with farmer groups to provide local language apps they want to use.
Legg adds that so far it distinguishes five major types of damage to cassava plants: three diseases and two types of pest damage.
Cassava virus diseases alone, explains Legg, cause losses of more than US$1 billion annually in Africa, and threaten food and income security of over 30 million farmers in East and Central Africa.
“The main target will be farmers in Sub-Saharan Africa. However, we will be working with the global network of CGIAR, and this means that the app could equally be of value in other parts of the developing world, such as Latin America and Asia.”
Peter Okoth, a consultant agronomist at the Kenya-based Newscape Agro Systems Ltd, tells SciDev.Net that smallholders in Africa cannot afford basic agricultural inputs, and thus well planned value chain arrangement with key players are needed to make its potential roll-out in Africa feasible.“For this app to generate the desired impact, the developers must partner with service providers and plant-health specialists and financiers to solve the problems,” explains Okoth. “The CGIAR needs to move a step further and constitute action consortia with membership drawn from an array of actors who are needed to address the practical aspects of solving the crop problems jointly with the farmers.”
Challenges in dissemination, according to Okoth, include information distribution and gaining potential users’ confidence that it will solve their problems as well as sustainability.