This story is part of “How Does It Work?” — The Better India’s series that takes the everyday things you see, use and rely on, and uncovers the clever technology, design and sustainable thinking that makes them work better. From the objects in your home to the systems around you, we break down the ideas behind them in simple, easy-to-understand language. Because innovation isn’t always about futuristic machines or complicated technology. Sometimes, it’s hiding in the things we use every day.
A leaf that once looked fresh and green can suddenly develop yellow patches, brown spots or a strange curling pattern. To the naked eye, it may simply look “unhealthy”. But for a farmer, identifying what caused the damage can mean the difference between saving a crop and watching a disease spread.
For a farmer, the first challenge is not just noticing that something is wrong — it is figuring out what is wrong. Is that yellowing caused by a fungal infection, an insect, a lack of nutrients or simply changing weather?
Traditionally, answering this may require close observation or expert help. But today, there is another tool that can offer a quick first clue: a smartphone camera paired with artificial intelligence.
So, how can a machine look at a photograph of a leaf and recognise something that may be invisible to an untrained eye? The answer lies in a technology called computer vision.
It starts with a photograph
The process begins with an image. A farmer can use a smartphone to photograph a leaf, stem or affected portion of a plant.
Computer vision analyses patterns such as colour, texture, spots and lesions to assess plant health.
Photograph: (Agrobit)
The clearer the photograph, the better. Good lighting, a close view and visible healthy and affected portions can give the AI more useful information to work with. In larger farms, cameras or drones can also capture images across fields.
But the AI isn’t simply “looking” at the photograph like a human would.
The AI looks for patterns
Once the image is uploaded, computer vision processes the pixels in it. The image may first be resized or adjusted for factors such as lighting so it can be analysed consistently.
Then comes the interesting part.
AI models learn to recognise crop diseases by training on thousands of labelled images of healthy and affected plants. Photograph: (Lets Nurture)
A deep-learning model, often a convolutional neural network (CNN), examines visual patterns such as colour, texture, shape, spots, lesions and where these appear on the leaf.
Think of it as teaching a computer to recognise patterns by showing it thousands of examples.
How does the AI learn what a disease looks like?
Before an AI tool can identify a disease, it has to be trained.
Developers feed the model large collections of labelled images — photographs marked as healthy or affected by particular diseases. By analysing these examples, the model learns which visual patterns are associated with different conditions.
When a new photograph is uploaded, the trained model analyses its patterns and compares them with what it has learned. It can then predict the most likely disease or condition.
The result isn’t necessarily limited to a disease name. Depending on the system, AI can also help identify the affected area, estimate the severity of damage and suggest what action may need to be considered next.
What happens after the diagnosis?
This is where the technology becomes particularly useful.
Instead of waiting for symptoms to spread across an entire field, farmers can use an early warning to inspect nearby plants, seek expert advice or take appropriate action. Repeated photographs can also help monitor whether the problem is improving or spreading.
Early detection can help farmers inspect nearby plants, seek expert advice and respond before a disease spreads further. Photograph: (Farmout)
However, AI isn’t a replacement for agricultural experts. A blurry photograph, poor lighting or symptoms that look similar across different diseases can affect its accuracy. And some infections may not show visible symptoms early enough to be detected from an ordinary photograph.
In such cases, field inspection, sensors or laboratory testing may still be needed.
Still, the idea is powerful: a simple photograph can become a source of information about crop health.
As AI models are trained on more crops, diseases and real-world field conditions, this technology could make crop monitoring faster and more accessible — helping farmers spot problems earlier, respond more precisely and give healthy plants a better chance to thrive.
