MLOps-Project

Project 1: Face Recognition

In this project, we focus on MLOps by developing a face recognition system. The objective is to allow users to enroll in the system by taking a photograph of their face. During subsequent logins, their faces will be verified before granting access. We will utilize a pre-trained model from the MTCNN library for this purpose.

User Instructions

• Create a new folder face-recognition-system
• Clone the GitHub repository git clone < paste the url here > cd face-recognition-system

Create a virtual environment

• Install the virtual environment: pip install virtualenv
• run: python -m < your-virtual-env-name >
• Activate your virtual environment: source < your-virtual-env-name >/bin/activate # (On windows use) < your-virtual-env-name >\Scripts\activate
• To deactivate the virtual environment, run: deactivate

• Run the requirements.txt file as it contains the Python dependencies that we will need for the project.
  • run: pip install -r requirements.txt
  • Run the main.py file: python main.py

Deployment

• API Development with FAST API
• In the main.py file, we have API endpoints for user enrollment, login, updates, deletions, and face recognition. Run the API using Uvicorn:
• uvicorn main:app --reload
• Copy this (http://127.0.0.1:8000/docs), open your browser and paste it.

Docker

• Ensure you have Docker installed, you can install it from here.

  • Build the docker image and run the docker container.

    docker build -t face-recognition-system . (copy and paste up to the dot)

    docker run -d -p 8000:8000 face-recognition-system

  • Open (http://localhost:8000) on your browser to access the FastAPI application.

Data Versioning

• Data versioning with DVC

  • Install DVC, if not yet installed.

    dvc init

  • Track data files assuming your dataset is in a folder named data.

    dvc add data/

  • Push data to a remote storage

    dvc remote add -d myremote <remote_storage_url> #the remote storage could be AWS, GCP, Azure etc

    dvc push

Frontend with Plotly Dash

• It provides an easy way to create a frontend interface for users to interact with the brain tumour segmentation service. We chose Plotly Dash as it is a powerful framework that's well-suited for data visualizations given that we will be using image data.

  • Create a plotly dash script, e.g app.py

  • Run the Plotly Dash application

    pip install dash

    python app.py

• After running the command, you will see an output ( http://127.0.0.1:8050/ or http://localhost:8050/ ).
• Open your browser and paste the URL to access the dash application.