AgroGuard

Inspiration

Farmers often face heavy crop losses because plant diseases are detected too late.
In many rural areas, language barriers and low literacy make it difficult for farmers to understand digital tools or written instructions. We wanted to build an AI-based system that helps farmers identify crop diseases early and receive guidance in their local language with voice support.

What it does

AgroGuard allows users to upload an image of a crop leaf and detects whether the plant is healthy or affected by diseases such as blight, rust, or leaf spot. The system provides:

• Disease detection using AI
• Clear guidance text
• Automatic voice instructions
• Support for multiple Indian languages
• A simple, farmer-friendly interface

How we built it

We trained a Convolutional Neural Network (CNN) using the PlantVillage dataset to classify crop diseases from leaf images. The system was built using:

• TensorFlow and Keras for model training
• Flask for the backend
• HTML, CSS, and JavaScript for the frontend
• gTTS for text-to-speech voice guidance

Challenges we ran into

Some diseases such as rust and leaf spot appear visually similar, especially in early stages.
Handling multiple languages with voice support while keeping the system simple was also challenging. Instead of forcing perfect accuracy, we focused on early detection and practical guidance.

What we learned

We learned how to build an end-to-end AI application and the importance of accessibility in real-world solutions.
This project taught us that honest and explainable AI is more valuable than claiming perfect accuracy.

What's next for AgroGuard

• Live camera-based disease detection
• Mobile application for farmers
• Improved accuracy using more real-world data
• Offline support for rural areas

Built With

• css
• dataset
• flask
• gtts
• html
• javascript
• keras
• plantvillage
• python
• tensorflow