Cascade-LBPH

Cascade Classifier

An application that records images, trains an algorithm for recognition of different faces registered, and detect in real time the face using a Cascade Classifier and LBPH Algorithm.

Dependencies

• Create a virtual environment, activate, and install the required libraries:

$ python -m venv .venv
$ source .venv/bin/activate # Unix
$ pip install opencv-contrib-python requests

How to use

Registering Faces

• To register new faces, run the script 01_face_registration.py:

$ python 01_face_registration.py --records="records" --id=1 --name="breno"

• Press 't' to record an image (record more than one image). Press 'q' to exit. This will create the following structured 📁 :

Cascade-LBPH/
├─ records/
│  ├─ 001/
│  │  ├─ img1.png
│  │  ├─ ...
│  ├─ id-name.json

• If the folder already exists, this will append the new ID in the folder

• NOTE: You need to close the application and rerun if you want to record another face (different ID)

Training the recognizer

• After saving the faces, it's time to train the recognizer with each ID using the images saved before 02_face_training.py.

$ python 02_face_training.py --records="records"

• In this step, will create two files: one for locate face (haarcascades/haarcascade_frontalface_default.xml) and the other one is the weights trained of the LBPH algorithm (trainer/trainer-YY-MM-DD-HH-MM-SS.yml)

Real-time detection

• Finally, it's time to detect the face and recognizer using the webcam in real-time with the code 03_face_detection.py

$  python 03_face_detection.py --records="records" --detector_filepath="haarcascades/haarcascade_frontalface_default.xml" --recognizer_filepath="trainer/trainer-2023-09-17-13-17-57.yml"