The Tanzania Project Shifts into High Gear

Dr. Neema Mduma and her colleagues at the Nelson Mandela African Institution of Science and Technology (NM-AIST) in Arusha, Tanzania have a dream: to build a powerful smartphone application that can help smallholder farmers detect the earliest signs of disease in staple crops.

Now, with the help of Grow Further’s donors, NM-AIST is getting closer to achieving this vision. Their innovation promises to be a game changer for farmers throughout sub-Saharan Africa who are constantly struggling with pests and diseases that are killing their crops.

“The data collection exercise, which began in earlier project phases, has now been completed in the Arusha, Kilimanjaro, and Manyara regions,” Dr. Mduma recently reported to us. “This marks a significant milestone.”

 

Kobo Collect

A little over a year ago, Grow Further proudly announced our grant to NM-AIST in support of an innovative research and development project. Mduma and her team proposed utilizing a powerful machine learning system to develop an app that can warn smallholder farmers of when their fields are threatened by diseases and pests. The system would analyze hundreds of thousands of images of all stages of disease progression or pest infestation and “learn” to detect these indicators when farmers take pictures of their crops’ leaves and stems.

Mduma and colleagues began by developing an image collection system. Then they spent months taking snapshots of parts of infected crops, collecting a huge treasure trove of data in the process. Mduma says their work has been made easier thanks to the help of more than 1,500 farmers enlisted in the effort, not to mention the five government agricultural extension officers who lent their hands.

“These upcoming activities will bring the project closer to its ultimate goal of providing smallholder farmers in Tanzania with AI tools for early detection of crop diseases.”

 

Their success is also thanks to Kobo Collect, an advanced data collection tool that they’ve used as the main conduit for gathering and cataloging images of maize and bean plants. As NM-AIST put it, Kobo Collect is a powerful and ideal tool to use because of its simplicity.

“Users with limited technological experience could quickly learn to operate it,” Mduma said in her report. Kobo Collect can also collect images and data offline, a necessity in much of rural Africa where internet connectivity often doesn’t exist.

“In some project areas, internet access is often unreliable or entirely unavailable,” Mduma confirmed. “Kobo Collect’s ability to function in offline mode became a critical asset in these conditions.”

In total, our partners at NM-AIST have so far collected more than 254,000 images of maize leaves, over 155,000 bean leaf images, and nearly 90,000 other images documenting all stages of plant disease progressions and pest infestations. Their goal was to achieve at least 500,000 images, and they’ve succeeded thanks to the support of Grow Further donors.

Now comes the next step

With the first two phases of the project largely out of the way, Mduma reported to us that NM-AIST is rolling out phase 3 of their plan. This phase entails “data preparation and model development” she explained.

Now that they have this abundance of visual data on crop diseases and pest indicators, they must organize it all for this data to be of any use in the smartphone application they’re working on.

“This will involve categorizing the images into the appropriate disease and health classes and conducting additional quality checks to ensure the dataset’s consistency,” Mduma wrote.

She admits that her team ran into some problems during the field data collection phase.

For starters, the weather didn’t always cooperate. They also had some difficulty finding classic signs of some crop disease outbreaks because those diseases aren’t necessarily prevalent during the season when field data collection occurred. Getting to some of the more remote villages also proved challenging logistically.

NM-AIST is anticipating running into some headwinds moving forward into phase 3 and beyond, as well.

Mduma noted that some farmers they reached out to were hesitant to get involved, fearful that the technology NM-AIST is developing posed some sort of threat to their farms and livelihoods. The researchers were able to overcome these fears with the help of government agriculture extension workers and local leaders. They must also continue to keep an eye on the weather.

This R&D project will require further fieldwork. Rain and mud slowed them down during the field data collection phase. Rain gear helped get them through the worst of Tanzania’s weather, and Mduma and the team anticipate that they’ll need to bring along raincoats, boots, and waterproof bags for their future field outings.

Abundant optimism

Despite the weather and some logistical challenges, so far NM-AIST is meeting all of its targets.

They’ve collected a massive haul of field data, hundreds of thousands of images of crop diseases. They’ve fostered valuable partnerships with government extension officers and more than 1,500 Tanzania smallholder farmers. They’re getting the word out.

Next, they plan to further refine and improve their dataset, organize it, and get it ready for the machine learning tool that will eventually power the app that they hope to deliver to millions of smallholder farmers. The researchers at NM-AIST Dr. Mduma are bursting with optimism and enthusiasm, as are we.

“These upcoming activities will bring the project closer to its ultimate goal of providing smallholder farmers in Tanzania with AI tools for early detection of crop diseases,” Mduma wrote, “contributing to improved food security and resilience against climate change.

 — Grow Further

Photo credit: Grow Further founder and CEO Dr. Peter Kelly speaks with Dr. Neema Mduma (seated to his right) and farmers in the field during his travels in Tanzania. True Vision Productions.   

Subscribe to Our Newsletter

Categories

Newsletter Signup

Newsletter Sign-up Popup