Inviting proposals to develop an AI tool for interpretation of Line Probe Assay (LPA) results.

Cotton Farming

AI-based early pest warning system

Today, large agriculture programs in India depend on manual counting of pests and present numerous challenges with identifying pests accurately and providing advice to the small-holder farmers. We are working on AI-driven technology which runs on a basic smartphone, and classifies and counts pests based on photos of pest traps taken by farmers and agriculture program workers. This solution can be used to provide millions of farmers with timely, localised advice, reducing crop loss and over-use of pesticides by improving the timing of usage.

The Problem

Cotton is one of the largest cash crops grown globally. It is India’s third largest crop after rice and wheat and 75% of it is grown by small-holder farmers who struggle with uncertainty in yield and income. One of the critical challenges they face is the inability to manage pests despite heavy usage of pesticides.

More than a billion people live in smallholder farmer households worldwide, and many of these farmers struggle with avoidable pest damage that can wipe out up to 50% of annual crop yield. For example, in India, for the 30 million people — 6 million farmers and their families — that depend on cotton farming for a living, inability to manage pests effectively is one the biggest risks. This, despite the fact that cotton accounts for close to half of India’s pesticide usage.

Using features generated from a convolutional neural network, our system first determines whether the image is valid. If it is, those features are then used to detect and count the number of pests that are present. Using this count and with the help of guidelines set by entomologists, the system recommends an action to the farmer—to spray or not to spray pesticide. In the cloud, the fully trained network is over 250Mb in size. Using a combination of network pruning and quantization, however, we have been able to compress the model to approximately 5MB, a size suitable to run on low-end smartphones in areas that do not have network connectivity. We train our models on a data set of around 4,000 images we built in collaboration with smallholder farmers all over the country.

Impact through AI

In Kharif 2020, we deployed our solution via an app through on-ground partners such as Welspun Foundation and Deshpande Foundation in 4 districts across 3 of the largest cotton-producing states in India – Gujarat, Maharashtra and Telangana. Through existing agricultural programs, we reached nearly 15,000 farmers who saw a benefit through increase in profit as well as a reduction in pesticide cost compared to 2019.

The Wadhwani Institute for Artificial Intelligence is one of 20 organizations that will share $25 million in grants from, credit and consulting from Google Cloud and coaching by Google’s AI experts as a grantee of the Google AI Impact Challenge. We are passionate about creating an impact for small-holder farmers in India and the developing world at large.

"Small farmers worldwide depend heavily on government and nonprofit programs to figure out what to do at every step of the crop cycle. By using AI to augment human capabilities and overcome systemic challenges in these large-scale programs, we can help millions of farmers. Pest management is just the beginning."

- Raghu Dharmaraju, VP, Products and Programs