Image Classifier with Flask API and Streamlit Frontend

This project demonstrates an Image Classifier using a Flask API for prediction and a Streamlit application as a user interface. The classifier uses a Random Forest model trained on the Iris dataset, and the predictions are made based on basic image features.

---

Project Architecture

Architecture Diagram

[Streamlit UI]
     |
     v
[Flask API]
     |
     v
[Random Forest Model (iris_classifier.joblib)]
     |
     v
[Prediction Returned to Streamlit]

1. Model Training: A RandomForestClassifier is trained using the Iris dataset and saved as a .joblib file.
2. Flask API: Exposes a /predict endpoint to classify uploaded images.
3. Streamlit Frontend: Provides an interactive UI to upload images and get predictions from the Flask API.
4. Dockerized Services: Flask and Streamlit are containerized using Docker for easy deployment.

---

Model Training

The Random Forest model is trained on the Iris dataset and saved as a joblib file in the joblib directory. Here is the code for training and saving the model:

---

Project Structure

image-classifier/
│
├── joblib/                      # Directory containing the .joblib file
│   └── iris_classifier.joblib
├── flask_app/                   # Flask application directory
│   ├── app.py                   # Flask API code
│   ├── Dockerfile               # Dockerfile for Flask
│   └── requirements.txt         # Dependencies for Flask
├── streamlit_app/               # Streamlit application directory
│   ├── app.py                   # Streamlit UI code
│   ├── Dockerfile               # Dockerfile for Streamlit
│   └── requirements.txt         # Dependencies for Streamlit
└── docker-compose.yml           # Docker Compose configuration

---

Docker Setup and Commands

Docker Compose Configuration

The project uses Docker Compose to manage the Flask and Streamlit services.

docker-compose.yml:

Docker Commands

• Build the containers:

  docker-compose build

• Start the services:

  docker-compose up

• Stop the services:

  docker-compose down

---

Control Flow

1. Model Training: A Random Forest model is trained and saved as a .joblib file.
2. Flask API: The API loads the .joblib file and exposes a /predict endpoint.
3. Streamlit Frontend: Users upload images, and the app sends the image to the Flask API for prediction.
4. Dockerized Services: Both Flask and Streamlit run as separate Docker containers.

---

Screenshots

Streamlit Interface

![Streamlit Interface]

Prediction Output

![Prediction Output]

---

How to Run the Project

1. Train the Model:

   python training_script.py

   This will generate the iris_classifier.joblib file in the joblib directory.

2. Build and Run the Docker Containers:

   docker-compose build
   docker-compose up

3. Access the Applications:

   • Flask API: http://127.0.0.1:5000
   • Streamlit App: http://127.0.0.1:8501

---

Future Improvements

• Train a custom model specifically for image classification.
• Deploy the service on cloud platforms like AWS or GCP.
• Add authentication to the API.

---

License

N/A