A new artificial intelligence (AI) app developed by researchers from Africa's Bioversity International led by Michael Selvaraj allows banana farmers to identify and combat outbreaks of pests and diseases. Designed to be used with a smartphone, both on and offline, it's claimed to operate with an average success rate of 90 percent.

The common banana is one of the world's most important crops with over 113 million tonnes harvested every year. Millions of people in the tropics depend on the slippery-peeled staple as an important part of their diet, and with 20 million tonnes grown for export, they are a major agricultural commodity.

Unfortunately, only a few species of bananas are widely grown and because they are essentially clones where every plant is almost genetically identical to the one next to it, they are uniformly vulnerable to pests and disease. That means that any outbreak of Panama disease, tropical race 4, black sigatoka, banana bunchy top virus, banana bacterial wilt, xanthomonas wilt, fusarium wilt, or black leaf streak can have a devastating effect, with the recent spread of tropical race 4 costing US$253.3 million in Taiwan alone.

To help overcome this, the Bioversity International team has taken advantage of the increasing access to smartphone networks in even remote areas to produce an AI app called Tumaini, which means "Hope" in Swahili. Installed in a phone or other mobile devices, it lets small farmers control banana pest and disease outbreaks.

"Farmers around the world struggle to defend their crops from pests and diseases," says Selvaraj. "There is very little data on banana pests and diseases for low-income countries, but an AI tool such as this one offers an opportunity to improve crop surveillance, fast-track control and mitigation efforts, and help farmers to prevent production losses."

According to the team, the Tumaini app is based on recent improvements in image-recognition technology and deep learning. It uses 20,000 uploaded images of banana diseases and pests to learn how to identify specific signs of infection or infestation, as well as which steps to take to counter the problem. As it does so, it records data from the diagnostic image, including its geographical location, to grow and refine the database.

Unlike previous computer detection systems, Tumaini works with more than detached leaves against plain backgrounds. Instead, it can zero in on different parts of the plant in situ while disregarding otherwise confusing background objects. It can also operate offline as well as link farmers to extension workers for help.

So far, Tumaini is in the beta phase with testing in Colombia, the Democratic Republic of the Congo, India, Benin, China, and Uganda, where it has shown a success rate of 90 percent. The goal is to one day create a satellite-powered, globally connected network for the general control of disease and pest outbreaks.

"This is not just an app," says Selvaraj. "But a tool that contributes to an early warning system that supports farmers directly, enabling better crop protection and development and decision-making to address food security."

The research was published in Plant Methods and the video below discusses the Tumaini app.

View gallery - 4 images