CropScan - AI-Powered Crop Disease Detection System

🌱 What it does

CropScan is an intelligent agricultural solution that uses deep learning to identify crop diseases from leaf images. Farmers and agricultural professionals can simply upload a photo of an affected crop leaf, and our AI model provides instant diagnosis along with detailed treatment and prevention recommendations.

The system can detect 18 different crop diseases including fungal, bacterial, viral infections, and pest infestations across multiple crop types, helping farmers take immediate action to protect their harvests.

How we built it

Frontend (React + TypeScript)

Framework: React with TypeScript for type safety
Styling: Tailwind CSS for modern, responsive design
UI Components: Custom-built components with Lucide React icons
State Management: React hooks for efficient state handling
Build Tool: Vite for fast development and optimized production builds

Backend (FastAPI + TensorFlow)

API Framework: FastAPI for high-performance REST API
AI/ML: Custom trained Keras/TensorFlow model using resnet50 architecture
Performance: Async/await patterns for optimal response times

Machine Learning Model

Architecture: Transfer learning with resnet50 base model
Input: 224x224 RGB images
Output: 18 disease classes + healthy classification
Training: Data augmentation with rotation, zoom, and flip transformations - Callbacks: Early stopping and model checkpointing for best performance

Cloud Infrastructure (Microsoft Azure)

Backend: Azure App Service for scalable API hosting
Frontend: Azure Static Web Apps for global CDN delivery
Environment: Production-ready with health checks and monitoring

🎯 Key Features

Disease Detection

18 Disease Types: anthracnose, gummosis, leaf miner, red rust, bacterial blight, brown spot, green mite, mosaic, fall armyworm, grasshopper, leaf beetle, leaf blight, leaf spot, streak virus, leaf curl, septoria leaf spot, verticillium wilt, plus healthy classification
High Accuracy: Trained on extensive crop disease datasets
Real-time Analysis: Instant results within seconds
Confidence Scoring: Shows prediction confidence percentage