faceDetection

Goal

Demonstrate face and eyes detection using Haar-cascade classifier in OpenCV.

Steps to detect face and eyes

1. Initialize opencv objects

let src = new cv.Mat(video.height, video.width, cv.CV_8UC4);
let gray = new cv.Mat();
let faces = new cv.RectVector();
let eyes = new cv.RectVector();
faceCascade = new cv.CascadeClassifier();
eyeCascade = new cv.CascadeClassifier();

2. Load classifiers

faceCascade.load('haarcascade_frontalface_default.xml');
eyeCascade.load('haarcascade_eye.xml');

haarcascade_frontalface_default.xml file is pre-trained classifier for face detection which uses Haar Cascade model. Similarly, 'haarcascade_eye.xml' file is pre-trained classifier for eye detection.

3. Detect faces

cap.read(src);
cv.cvtColor(src, gray, cv.COLOR_RGBA2GRAY, 0);
faceCascade.detectMultiScale(gray, faces, 1.1, 3);

OpenCV function detectMultiScale(...) detects faces. Arguments of the function:

• gray is a matrix of the type CV_8U containing an image where objects are detected.
• faces is a vector of rectangles where each rectangle contains the detected object. The rectangles may be partially outside the original image.
• 1.1 is a scaleFactor specifying how much the image size is reduced at each image scale.
• 3 is a parameter specifying how many neighbors each candidate rectangle should have to retain it.

4. Detect eyes

For each face we copy face rectangle and pass it to eye detection classifier:

let faceGray = gray.roi(face);
eyeCascade.detectMultiScale(faceGray, eyes, 1.1, 3);
faceGray.delete();

References

1. Haar-cascade face and eyes detection in OpenCV
2. Face Detection Camera Example (OpenCV)