---

Eczema Detection and Classification Using Deep Learning

Project Overview

This project involves training a deep learning model to detect and classify different types of eczema using TensorFlow. The model is based on the InceptionResNetV2 architecture and includes a custom layer for scaling inputs. The final model is integrated into a Flask backend for predictions, which can be accessed via a React frontend.

File Structure

.
├── backend/
│   ├── app.py                # Flask backend for serving predictions
│   ├── final_model.h5        # Pre-trained deep learning model (not included)
│   ├── training.ipynb        # Jupyter notebook for training the model
│   ├── test/                 # Folder containing the eczema dataset
│   │   ├── Atopic Dermatitis/
│   │   ├── Contact Dermatitis/
│   │   ├── Dyshidrotic Eczema/
│   │   ├── Nummular Dermatitis/
│   │   ├── Seborrheic Dermatitis/
│   │   └── Stasis Dermatitis/
└── frontend/
    ├── public/               # React frontend public assets
    ├── src/                  # React frontend source code
    ├── package.json          # Node.js dependencies for React
    └── README.md             # Instructions for setting up the frontend

Prerequisites

Ensure you have the following software installed:

• Python 3.x (3.8 or 3.9)
• TensorFlow 2.x (for GPU support)
• Node.js and npm for frontend development
• Flask for the backend
• Pillow for image processing
• NumPy for numerical operations
• Scikit-Learn for data manipulation

Step-by-Step Setup and Execution

1. Clone the Repository

git clone https://github.com/yourusername/eczema-detection.git
cd eczema-detection

2. Backend Setup

Install Python Dependencies

Ensure you have a Python 3.x environment with TensorFlow version 2.x for GPU support. Install the required Python packages:

pip install tensorflow==2.9.1 flask pillow numpy scikit-learn matplotlib

Prepare the Dataset

Download the eczema dataset from Kaggle and extract it into the backend/test/ directory.

Train the Model

Run the training.ipynb notebook to train the model. This notebook will preprocess the images, train the model, and save it as final_model.h5.

Note: The final_model.h5 is not included in the repository. You will need to train the model yourself using the provided notebook.

Run the Flask Backend

After training, you can start the Flask backend using the following command:

cd backend
python app.py

The server will start on http://0.0.0.0:5000. You can send POST requests to /predict with image files for predictions.

3. Frontend Setup

Install Node.js Dependencies

Navigate to the frontend directory and install the required Node.js packages:

cd ../frontend
npm install

Run the React Frontend

After installing the dependencies, you can start the React frontend by running:

npm start

This will start the development server on http://localhost:3000, which will interact with the Flask backend for predictions.

Using the Application

1. Start the Flask Backend:

   cd backend
   python app.py

2. Start the React Frontend:

   cd ../frontend
   npm start

3. Interact with the Application: Open your web browser and navigate to http://localhost:3000 to use the frontend, which will send images to the Flask backend for eczema detection.

Troubleshooting

Custom Layer Error

If you encounter an error related to the custom layer CustomScaleLayer, ensure that:

• The class is defined in the app.py file.
• The load_model function in app.py includes custom_objects={'CustomScaleLayer': CustomScaleLayer}.

TensorFlow GPU Support

This project uses TensorFlow version 2.x for GPU support. Ensure you have the correct version installed by running:

pip install tensorflow-gpu==2.x

Model Not Included

The pretrained model (final_model.h5) is not included in the repository. You need to train the model using the provided notebook before running the backend.

Conclusion

This project provides a full-stack solution for detecting and classifying eczema using a deep learning model, Flask for backend API, and React for frontend UI. Follow the setup instructions to train the model and deploy the application.

---