Plant Disease Detection using CNN

This project is a high-accuracy plant disease detection system deployed as a full-stack web application. It uses a PyTorch-based Convolutional Neural Network (CNN), built on a ResNet50 architecture, to classify 38 different plant diseases from leaf images. The application is served through a Python (FastAPI) backend API and consumed by a modern React frontend.

Model Details

The model was trained using transfer learning on a pre-trained ResNet50 architecture.

• Framework: PyTorch
• Base Architecture: ResNet50 (from torchvision.models)
• Performance: Achieved 97.7% validation accuracy
• Key Techniques:
  • Transfer learning (only the final fully-connected layer was trained).
  • Data Augmentation (RandomFlip, Rotation, ColorJitter) to improve robustness.
  • Class Balancing using WeightedRandomSampler to handle imbalanced data.
• Trained Model: The best model weights are saved in plant_disease_resnet50_best.pth.

Dataset

The model was trained on the New Plant Diseases Dataset (Augmented), which is publicly available on Kaggle. This dataset contains over 87,000 RGB images of plant leaves, categorized into 38 distinct classes (including healthy and diseased leaves).

• Source: Kaggle: New Plant Diseases Dataset (Augmented)
• Classes: 38

Project Structure

The project is now divided into three main components:

• /model_training.ipynb: The Jupyter notebook containing all steps for data loading, preprocessing, model training, and evaluation.
• /plant-disease-api/: The backend Python API (FastAPI) that serves the trained PyTorch model.
• /plant-disease-ui/: The new frontend web application built with React.

How to Run the New Application

To run the complete application, you must start both the backend API and the frontend UI.

1. Backend (API)

(Instructions based on the plant-disease-api folder)

# Navigate to the API directory
cd plant-disease-api

# Install required Python packages
pip install -r requirements.txt

# Run the API server (using uvicorn)
uvicorn app.main:app --reload

2. Frontend (React App)

(Instructions from the plant-disease-ui/README.md)

In a new terminal:

# Navigate to the UI directory
cd plant-disease-ui

# Install node modules
npm install

# Run the app in development mode
npm start

Once both services are running, you can view the app in your browser at http://localhost:3000.