Twenty to 40% of global crop production is lost to pest and disease damage, costing more than US$200 billion every year. Consequently, one of the top priorities of agricultural researchers worldwide is finding ways to identify, monitor, and control crop pests and diseases.
The first step in this process is identification. People have probably been able to recognize some of the most apparent pests and diseases since the earliest days of farming. Pestilence from locusts is described in many of the ancient religious texts. Causal agents of diseases, however, by their microscopic nature, have only been accurately distinguished in more recent times. The development of microscopy and biochemical assays delivered large-scale improvements in the accuracy of pathogen identification, and from the late 20th century, the DNA revolution allowed for the development of diagnostic tests that could identify any organism. Nevertheless, most farmers and farm workers continue to use visual symptoms or appearance to identify crop pests and diseases, mainly since accurate diagnostics are either too expensive for farmers or inaccessible.
Mobile phone technology has spread rapidly across the globe over the last 30 years, and current expansion is most rapid in Africa. The coupling of phones with the internet and high spec cameras has also now offered unique new opportunities for these portable devices to be used as mobile pest/disease diagnostic units.
In 2016, IITA began working with the PlantVillage team of David Hughes, at Pennsylvania State University, USA, on innovative technologies for diagnosing cassava pests and diseases. It quickly became clear that the best bet for a diagnostic that could support millions of farmers directly was a mobile phone-based system that used artificial intelligence (AI).
Thousands of images were collected of the most crucial pest/disease symptom types, including cassava mosaic disease (CMD), cassava brown streak disease (CBSD), cassava green mite damage as well as symptom-free healthy leaves. The images were then annotated by IITA pest/disease experts to indicate the leaf portions showing characteristic symptoms. The Penn State team then applied machine learning techniques to train computers to recognize each of the symptom types.
The final step was to build these AI models into a smartphone app, which would prompt the user to point the phone’s camera at cassava leaves and then deliver an instant diagnosis of the disease or pest damage. The app was called PlantVillageNuru, with ‘Nuru’ being the Swahili word for light.
Users of PlantVillageNuru could download the app when online, but a key quality of the new app was that once installed, it could be used offline, allowing farmers in their fields anywhere to make diagnoses. PlantVillageNuru was made available for free download on Google’s PlayStore in June 2018. In addition to providing instant pest/disease diagnosis, the app also provides guidance on control, and helps farmers to source healthy planting material of improved varieties by linking users to the IITA SeedTracker app.
As PlantVillageNuru was made available to farmers worldwide, it was important to compare its performance with likely users and to use this comparison to measure improvements in the app over time. To achieve this, a set of 170 images of cassava leaves was assembled, with different pest/disease damage symptoms or no symptoms at all. A team of experts confirmed the diagnoses of each of the images, and these were assembled into a PowerPoint presentation referred to as the Cassava Symptom Recognition Assessment Tool (CaSRAT). IITA researchers – Latifa Mrisho and Neema Mbilinyi – then set out to use the tool to see how well Nuru compared in its diagnoses to different groups of people, including researchers, extensionists, and farmers in coastal Tanzania, either trained or untrained.
Unsurprisingly, they found that trained researchers had the highest levels of correct diagnosis (85%), followed by trained extensionists (49%), while groups with the lowest scores were untrained farmers (16%) and untrained researchers (21%). Although the accuracy of Nuru was only 40% in 2018, subsequent upgrades have increased this to 65% by 2020, which is higher than all other groups apart from the trained researchers. Further work by Latifa and colleagues in western Kenya has also shown that using PlantVillageNuru over a period of just one week can lead to improvements in pest/disease recognition by farmers and extensionists, highlighting the value of the app as a training tool.
PlantVillageNuru has already been used by farmers from across Africa to generate more than 15,000 reports. Fortunately, millions of cassava farmers have access to smartphones, so there is huge potential for further scaling. Penn State and IITA are therefore partnering with a range of other initiatives to drive this process forward. A BigData Platform, CGIAR-Inspire scaling grant, is supporting further training and awareness raising in five countries of East and Central Africa, and PlantVillage is working with the WAVE and Technologies for African Agricultural Transformation (TAAT) projects to scale Nuru use in 11 countries of West and Central Africa. A partnership with CGIAR’s Roots, Tubers and Bananas (RTB) Program is also facilitating the application of the approach to other RTB crops, such as potato and sweet potato. These developments give great encouragement to the team working on Nuru that millions of farmers will soon have the power of pest and disease diagnostics in their hands.
Latifa recalls from her experience in introducing the app, “This tool offered the means to tackle challenges that farmers face in growing crops and the opportunity to learn how to improve productivity. Having access to information they need to improve their productivity gives farmers and agricultural officers power and control over their livelihoods.” It is clear, therefore, that this smart app is set to play a vitally important role in transforming African agriculture in the coming years.
Authors: James Legg, Juma Yabeja, IITA-Tanzania; Lava Kumar, IITA-Nigeria (Ibadan); Regina Kapinga, Edward Kanju, IITA-Uganda;
Silver Tumwegamire, IITA-Rwanda; Rudolph Shirima, Gloria Ceasar, Hekima Mtoji, IITA-Tanzania; Busayo Ogunya, (Ibadan); George
Swella, Tanzania Official Seed Certification Institute; Kiddo Mtunda, Heneriko Kulembeka, Tanzania Agricultural Research Institute;
Stephen Magige, David Eagle, Mennonite Economic Development Associates.