VVG Face Identification

Author: Rustenis Tolpeznikas Project year: 2024

This repository contains a single Jupyter notebook that demonstrates a webcam-based face identification workflow. It uses MTCNN for face detection, OpenCV for webcam and image handling, and VGGFace embeddings to compare a reference face against live captures.

What It Does

• Captures frames from a local webcam.
• Detects faces in real time with MTCNN.
• Builds a reference face embedding from a captured face.
• Compares subsequent faces against the reference embedding.
• Draws bounding boxes to distinguish reference and verified faces.

Requirements

Install the Python packages used by the notebook:

pip install opencv-python mtcnn numpy keras_vggface scipy

You will also need:

• Python 3.x
• Jupyter Notebook or JupyterLab
• A working webcam
• A local desktop session, since OpenCV GUI windows are used

Usage

1. Open Notebook.ipynb.
2. Run the notebook cells from top to bottom.
3. Capture a reference face when prompted by the notebook logic.
4. Trigger verification to compare live faces against the stored reference.

Notes

• The notebook is intended as a lightweight demo and starting point rather than a production-ready recognition pipeline.
• Webcam access and OpenCV display windows may not work in remote or headless environments.
• If you extend the notebook, keep the face-cropping and embedding steps aligned so the comparison remains meaningful.

License

Licensed under the Apache License 2.0. See LICENSE for details.