• 

  CropScan AI

Inspiration

As a student from a rural community, I’ve seen firsthand how farmers struggle to identify and treat crop diseases early. Many don’t have access to agricultural experts or reliable tools for diagnosis. This inspired me to create CropScan AI — a simple, AI-powered web tool that allows farmers to detect diseases from photos of leaves, using only their phone.

What it does

How we built it

The project uses a convolutional neural network (CNN) trained on a subset of the PlantVillage dataset. I built the frontend using Streamlit, allowing users to upload leaf images. The backend processes the image, makes a prediction, and returns both the diagnosis and actionable advice.

Simple simulated inference

if "spots" in uploaded_image: prediction = "Tomato Leaf Spot" The system is modular and ready for future integration with real-time APIs, SMS alerts, or local language translation for wide accessibility.

Challenges we ran into

Limited access to hardware for full model training, so I simulated outputs during early prototyping

No funding to deploy the app at scale, so I optimized for local use

Working solo meant I had to wear many hats — from ML engineer to UI designer to video editor

Despite these, I stayed focused on building a functional, impactful, and demo-ready tool to empower local farmers.

Accomplishments that we're proud of

What we learned

Throughout this project, I learned how to:

.Preprocess and classify image data using TensorFlow and Keras .Build an intuitive user interface using Streamlit .Simulate AI predictions to create functional MVPs .Organize code, create a clean GitHub repo, and write effective README files .Prepare submission materials like demo videos and pitch scripts

What's next for CropScan AI

Built With

• python
• streamlit
• tensorflow