🌸 Flower Disease Recognition using Deep Learning

📌 Project Overview

This project builds an end-to-end flower recognition and disease detection system using deep learning.
Given an input image, the model predicts:

• Flower species (lily, rose, sunflower)
• Health status (healthy or diseased)

The system is designed for quick screening and early disease detection, reducing reliance on manual inspection.

⚙️ Data & Preprocessing

• Image dataset organized by species and health condition
• Invalid and corrupted images removed
• Images standardized to RGB format and resized
• Targeted data augmentation applied only to diseased images to address class imbalance

🧠 Modeling Approach

• Transfer Learning with a pre-trained ResNet50V2 backbone
• Multi-task learning architecture:
  • Shared CNN feature extractor
  • Two output heads:
    • Species classification (3 classes)
    • Disease detection (2 classes)
• Two-phase training:
  • Frozen backbone training
  • Fine-tuning top layers with a low learning rate

📊 Results

• Species classification accuracy: ~84%
• Disease detection accuracy: ~98%
• Strong recall for diseased class, suitable for screening use cases
• Transfer learning significantly outperformed custom CNN baselines

🚀 Deployment

This project is a full-stack app with:

• backend/ running a Flask API and serving the trained Keras model
• client/ running a React frontend that uploads images to the backend

Run locally

1. Backend
   • cd backend
   • python -m venv .venv
   • .venv\\Scripts\\activate
   • pip install -r requirements.txt
   • python app.py
2. Frontend
   • cd client
   • npm install
   • npm start

The frontend uses the package.json proxy http://localhost:5000/ and the backend defaults to port 5000, so local requests should work without additional configuration.

If you want to point the frontend to a different backend URL, set REACT_APP_API_URL=http://your-backend-url before running npm start.

Useful links

• GitHub: https://github.com/Raj-Rathod-Ai/FlowerDiseaseSystem
• LinkedIn: https://linkedin.com/in/raj-rathod-ai

Recommended free deployment

For this stack, the best approach is:

• Railway: easiest free option for hosting the Flask backend and static React app in separate services
• Render: also supports a free Flask service, with the React frontend deployed separately as static sites

Suggested deployment flow

1. Deploy backend/ on Railway or Render as a Python app.
2. Build client/ with npm run build and deploy the generated static site on Vercel, Netlify, or Render.
3. Update the frontend API URL if needed to point at the deployed backend.

Why these platforms?

• Railway and Render support Flask apps directly on free tiers.
• Vercel/Netlify are ideal for hosting the React frontend as static files.
• This setup is the cleanest fit for your current client + backend repo structure.

📁 Folder Structure

Flower-Disease-Detection-using-Deep-Learning-main/
├── backend/          # Flask API and model serving
├── client/           # React frontend application
├── manifests/        # label map and dataset metadata files
├── models/           # trained Keras model files
└── README.md         # project documentation

Folder details

• backend/: contains the Flask app and API endpoint for image upload and prediction.
• client/: houses the React app, UI components, and CSS styling for the web interface.
• manifests/: stores supporting JSON and CSV files used for label mapping and dataset organization.
• models/: holds saved model weights used by the backend for inference.

🛠️ Tech Stack

• Programming Language: Python
• Deep Learning Framework: TensorFlow / Keras
• Image Processing: OpenCV, Albumentations
• Backend / API: Flask
• Frontend: React
• Development & Experimentation: Jupyter Notebook

📌 Key Takeaway

This project demonstrates how deep learning and transfer learning can be applied to real-world image classification problems, handling class imbalance, enabling multi-task predictions, and supporting deployable ML systems.