Face Recognition Pipeline (MediaPipe + LBPH)

A lightweight and practical face-recognition system combining:

• MediaPipe Face Mesh for face detection
• OpenCV LBPH for classical face recognition

This project provides a complete workflow: capture → train → recognize.

---

Project Structure

project/ │── capture.py # Capture and store face images │── train.py # Train the LBPH model │── predict.py # Real-time face recognition │── dataset/ # Auto-created user image folders │── models/ │ ├── lbph_model.xml │ └── label_map.json │── README.md

---

1. Capture Face Images

Run:

python capture.py

You will be asked to enter your name. Look into the camera and press Q to stop.

Images are saved in:

dataset/<your_name>/

---

2. Train the LBPH Model

Run:

python train.py

This generates:

models/lbph_model.xml models/label_map.json

---

3. Real-Time Face Recognition

Run:

python predict.py

The webcam window will display:

• A green rectangle around the face
• The predicted name
• The LBPH confidence score

Press Q to exit.

---

Installation

pip install opencv-python mediapipe pip install opencv-contrib-python

---

How to Use

git clone https://github.com/amani-patrick/Facial_recognition.git

cd Facial_recognition

python -m pip install mediapipe opencv-python

Then run:

1. python capture.py
2. python train.py
3. python predict.py

---

Notes

• Works with any number of people.
• For each new user, repeat capture → train.