Cropped AI

Disease treatment screen
Home screen
Prediction results screen 1
Prediction results screen 2
Finding a treatment supplier

Inspiration

We were inspired by the challenges smallholder farmers face in identifying and treating crop diseases early. Many lack access to agricultural experts or timely diagnosis, leading to major crop losses. We wanted to build something that could put expert-level help right in their pockets, and that’s where Cropped AI was born.

What it does

Cropped AI lets users:

Snap a photo of a diseased leaf
Instantly get a diagnosis using a trained AI model
Receive actionable treatment recommendations
See nearby suppliers that stock the needed products

It’s designed to be fast, simple, and useful in real-world farming conditions.

How we built it

We trained a MobileNetV2-based image classification model using a dataset of crop leaf images. The backend was built with FastAPI and deployed to Render. It exposes a prediction API that the mobile app, built using React Native with Expo, calls in real-time. For supplier data, we use OpenStreetMap queries based on the user’s location.

Challenges we ran into

Model training was slow since we used our personal laptops
Render backend sleeps after inactivity, causing delays during testing and affecting the experience during low usage periods.
Working with OpenStreetMap required trial-and-error to fine-tune queries for relevant agricultural suppliers.