CV module: Face Extractor

Crops faces out of the webcam image and saves them into the directories data/training and data/validation to train a classifier.

Algorithm description

1. The algorithm asks for a label (name) and
2. makes a new directory in data/training and data/validation with the scheme NAME.HASH which contains an unique hash value to distinguish similar or equal labels from each other.
3. Then it activates the webcam and makes 50 pictures every 100ms of the users face in front of it and saves them into the newly created folder with the pattern NAME.HASH(Timestamp).jpg. When no face is detected, no picture will be saved.
4. Secondly, the user has to hit the enter key to save another 50 pictures as validation set. These will be saved into data/validation.

Requirements

Face detector algorithm needed, that returns faces, eyes and image.

List of possible detectors:

• Face Detector module by @n1g1r1

Usage

Import the module, as well as the detector in your file and call the method like the following:

face_extractor.build_training_set(detector, classifier = "lbp")

• detector is the detector module you've chosen.
• classifier is the chosen classifier. Default: lbp.

Example

In this example the module structure is like the following:

└── modules
    ├── face_extractor
    └── face_detector

First, import the modules:

from modules.face_extractor import face_extractor as extractor
from modules.face_detector import face_detector as detector

Then, call the build_training_set method and set the detector as the used detector method:

extractor.build_training_set(detector)