🌿MangoMediX

AI-Powered Mango Leaf Disease Prediction System

MangoMediX is a deep learning–based web application that predicts mango leaf diseases from uploaded images.
With an accuracy of 92%, it empowers farmers and researchers with real-time diagnosis and actionable treatment recommendations, reducing reliance on manual inspections.

---

✨ Features

• 📸 Upload mango leaf images via a clean web interface
• 🤖 Predicts 8 mango leaf conditions:
  • Anthracnose
  • Bacterial Canker
  • Cutting Weevil
  • Die Back
  • Gall Midge
  • Healthy
  • Powdery Mildew
  • Sooty Mould
• 📊 Displays prediction confidence score
• 💡 Provides disease-specific treatment suggestions
• 🖥️ User-friendly, responsive design
• 🧠 Model built on ResNet50 transfer learning with TensorFlow/Keras

---

🛠️ Tech Stack

• Frontend: HTML, CSS, JavaScript
• Backend: Python, Flask, Flask-CORS
• Deep Learning: TensorFlow, Keras, ResNet50 (Transfer Learning)
• Storage/Reports: JSON (training history & evaluation metrics)

---

⚙️ Setup Instructions

1. Clone the repository

   • git clone https://github.com/vignesh-p3007/MangoMedix.git
   • cd MangoMedix

2. Create a virtual environment (recommended)

   • python -m venv venv
   • source venv/bin/activate # for Linux/Mac
   • .\venv\Scripts\activate # for Windows

3. Install dependencies

   • pip install -r requirements.txt

4. Run the web app

   • python backend/app.py

5. Access in browser

   • http://127.0.0.1:5000/

---

Screenshots

Here are some screenshots of the MangoMedix web application:

1. Home Page

2. Upload Leaf Image

3. Prediction Result

---

Pre-trained Model

The trained model disease_detector.h5 (~271 MB) is hosted externally due to GitHub file size limits.

Download the model here: Google Drive Link

Instructions:

1. Download the disease_detector.h5 file from the link above.
2. Place the file in the backend/ folder of the MangoMedix project.
3. Run the web application as usual:

python backend/app.py

---

Demo

Watch the MangoMedix web app in action:

Demo Video: Click Here to View

---

🚀 Live Demo

Check out the live demo of MangoMediX here:
👉 Click to View

---

📊 Results & Accuracy

• Achieved 92% prediction accuracy using ResNet50 with transfer learning
• Early detection helps prevent disease spread and optimize resource usage
• Scalable to larger datasets and adaptable for other crops

---

🚀 Future Improvements

• Expand dataset to include more diseases
• Mobile-friendly version for farmers
• Deploy on cloud (Render, Hugging Face Spaces, or AWS)
• Add notifications for disease alerts
• Multilingual support

---

📜 License

This project is licensed under the MIT License.

---

👨‍💻 Authors

• Vignesh
• Kshitij R Amin
• Nitheesh Ishwar Naik
• Suhas S

Guided by: Mrs. Usha C S, Assistant Professor, AJIET Mangalore