B3 · Applications of Artificial Intelligence

AI in Agriculture

AI in agriculture helps detect crop disease from photos, forecast yields and weather, guide irrigation and fertiliser, and give advice in local languages. For small farmers, cost, connectivity and language are the main barriers.

8 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

AI in agriculture helps detect crop disease from photos, forecast yields and weather, guide irrigation and fertiliser, and give advice in local languages. For small farmers, cost, connectivity and language are the main barriers.

In simple words

A farmer photographs a sick leaf on a phone, and an app names the likely disease and what to do, in Marathi or Hindi. The same idea as photo recognition, applied where it can protect a family's income.

Why it matters

Agriculture is a very large part of India's workforce, so useful tools here could help many people. It also shows that a model needs data from local crops and conditions.

Key terms

Precision farming
Giving each part of a field exactly the water, fertiliser or spray it needs, and no more.
Crop advisory
Advice to farmers on when to sow, water, spray and harvest.
Yield
The amount of crop harvested from a given area of land.
Remote sensing
Measuring the land from a distance, using satellites or drones.
Smallholder
A farmer who works a small piece of land, as most farmers in India do.
Pest
An insect or other creature that damages crops.

Explained in detail

Farming is full of predictions

A farmer makes predictions every week. Will the rain come on time? Is this the right day to sow? Are those spots on the leaf the start of a disease? How much will the crop sell for in two months? For centuries farmers answered from experience and from the advice of neighbours. AI adds a new source: patterns found in weather records, satellite images, and photos and reports from millions of other fields.

What AI can do on a farm

TaskHow it worksBenefit
Naming a diseaseThe farmer photographs a leaf; a vision model compares it with thousands of labelled leaf photosEarly treatment with the right remedy
Warning of pestsCounts of insects caught in traps, with weather data, predict an outbreakSpraying only when needed
Advice on sowingModels of rainfall and soil moisture suggest the best week to sowFewer failed sowings
Watching fields from aboveSatellite images show which fields are stressed or damagedFaster help and insurance claims
Spraying preciselyA camera on a sprayer tells weeds from crop and sprays only the weedsLess chemical, lower cost

Examples from India

Several projects in India show what is possible. In Andhra Pradesh, a sowing-advice service that sent farmers text messages based on rainfall and soil data reported clearly higher yields for the groundnut farmers who followed it. For cotton, an Indian non-profit organisation built a phone tool that counts pests from a photo of a trap and advises whether to spray. The government's Kisan e-Mitra chatbot answers questions about farm schemes in several Indian languages, by text or voice. Crop insurance programmes use satellite images to estimate losses after floods and droughts.

The barriers are not technical

Most Indian farmers are smallholders, working plots smaller than two hectares. For them the hard part is rarely the AI model. It is everything around it.

  • Language: advice must come in the farmer's own language, and preferably by voice.
  • Connection: many fields have weak mobile signal, so tools must work with little data.
  • Cost: a tool that needs an expensive phone or a paid subscription excludes the poorest.
  • Local fit: a model trained on crops in one state may be wrong for the soil and varieties of another.
  • Trust: a wrong recommendation can cost a season's income, so farmers rightly want proof.

How it works

UseTechnique
Disease detectionImage classification
Yield forecastsRegression, satellite data
Advice in local languagesLLMs, speech
Pest and weather alertsTime series

Try it yourself

From a sick leaf to an answer

Follow a farmer's photo as it travels.

From a sick leaf to useful advice

A farmer sees a sick leaf. Put the steps of the app in the right order.

  1. 1The model compares the photo with the patterns it learned
  2. 2The farmer takes a photo of the leaf on a phone
  3. 3Long before this, the model learns from many labelled photos of healthy and sick leaves
  4. 4The app names the likely disease
  5. 5The app gives advice in the farmer's own language
  6. 6The farmer decides what to do, and asks an expert if unsure

Common mistakes

  • Using a model trained on foreign crops or lab photos in real Indian fields.
  • Designing for smartphones with fast internet only, when many users have neither.

Check yourself

Question 1

1. Detecting disease from leaf photos uses:

Question 2

2. Why might a good lab model fail in fields?

Summary

  • AI helps with disease, yields and advice.
  • Local data and languages decide whether it works.

Finished reading?