Indian startup collaborates with Google.org to deploy Artificial intelligence to prevent crop damage

Agriculture supports 70% of rural households in India, with 80 percent of smallholder farmers (those with two hectares or less) accounting for the majority . Cotton, India's third-largest agricultural output after rice and wheat, is cultivated by approximately 6 million people. It's prone to pests in particular: A particularly ravenous-type insect killed half of Maharashtra's cotton crop in 2017, the country's second-most populated state.

Wadhwani AI, an Indian non-profit firm, has partnered up with Google's philanthropic arm in its first endeavor in Asia to assist farmers avoid such tragedies and thereby enhance crop yields and revenues.

They’ve worked together to re-launch the nonprofit's "AI-powered farm decision support system," which is an app that uses artificial intelligence to identify the presence of the Pink bollworm pest that can severely damage cotton crops, predict the disease's trajectory, and receive practical cure suggestions from agricultural specialists. They hope that by doing so, they would be able to protect farmers' livelihoods and reduce the widespread use of pesticides, which can cause more harm than good.

Wadhwani AI is a non-profit organisation that focuses on using artificial intelligence for social good. Its goal with this project is to assist cotton farmers in gaining more revenue stability in a field that often delivers little. Pests like the pink bollworm are expected to kill 20 percent to 30 percent of Indian cotton crops on an annual basis, adding to the uncertainty.

The insects deposit their eggs on the cotton bolls, or seed capsules, and the larvae devour the seeds and destroy the fibers, lowering the amount and quality of the harvest. They're difficult to spot visually, but having a means to spot them before they enter the boll could save farmers' lives.

Smallholder farmers can use the CottonAce app to simply take a photo of pests collected in pheromone traps; the app then checks the image's authenticity before classifying and counting the pests. According to Dhruvin Vora, senior product manager at Wadhwani AI, the Artificial Intelligence model, which was trained on over 30,000 images, forecasts "the next generation of eggs and larvae before they become an infestation for the crop."

The AI model is embedded into the android app, which means that the pest identification and advice happen almost instantly on the phone itself, allowing the app to work in areas with limited connectivity. Field staff are available to provide assistance in order to help village farmers, but they may not be able to reach the farthest-flung farms in time to save the harvests; yet, they "have the community's trust baked in," according to Vora. The app is available through partners such as farmer welfare programs, who will also have access to a real-time dashboard through the app, which will keep them updated on treatment progress.

According to Vora, each geographic community has one or two "lead farmers," who are more tech-savvy and educated members of the community; all they'd need is the app (which can work without an internet connection), and they'd be able to relay the information to the 10 to 40 other farmers in their area at village meetings. In most cases, the suggestion will boil down to whether or not to spray.

If pesticides are required, the app will instruct farmers on how to spray in a scientifically sound manner, which is preferable to the indiscriminate, widespread use of general pesticides, which harms farmers' health, soil quality, and the environment, all while draining earnings. “Farmers who are already struggling with poverty are pushed even further into despair,” adds Vora. They may offer organic treatments such as neem oil, a biopesticide derived from India's neem tree that has pest-repelling characteristics, for less serious infestations.

Last year, Google launched a $10 billion investment to assist India's small companies modernize, including artificial intelligence for social good in agriculture, which CEO Sundar Pichai described as a "very personal" endeavor. Wadhwani AI was also selected from thousands of applicants as one of just a few companies, and the first cohort of Asian winners, to participate in a Google.org fellowship program, receiving a $2 million grant and the assistance of a group of nine fellows.

Over the course of six months, they worked with Wadhwani AI's team to relaunch the app, conducting more in-depth UX research, a full app redesign, and a new infrastructure to speed up the AI learning model. all of which will benefit Wadhwani AI’s other social-good projects, such as its baby-weighing machine.

CottonAce pilots were deployed in Maharashtra, Gujarat, and Telangana from June to December, during India's autumn harvest. According to an impartial review, smallholder farmers who followed the proposed guidance saw up to a 26.5 percent rise in profit margins and a 38 percent reduction in pesticide expenses. The business plans to launch the app in its entirety during this year's harvest, with the objective of reaching 300,000 farmers this year and a total of 2 million farmers by 2022.

Of course, most of this work has had to be done in India, where an unusually high number of COVID-19 deaths have occurred, as well as the appearance of a novel coronavirus variety. According to Robert Tung, Google.org product manager and one of the fellows, the limited amount of on-the-ground work had to be transferred remotely with the help of brand-new tools that the team "hacked together."

“Throughout it all, I've been encouraged by the farmers' tenacity in the face of the pandemic,” he says. “We listened and witnessed firsthand as they battled the pandemic with the same tenacity and bravery they bring to every obstacle they face,” Vora continues.