Face Detection

Detect faces in a live video feed with CVZone, and use their position and size in your own code

Face detection

Face detection is the process of finding faces in images or video and working out where they are. (It doesn’t tell you who someone is: that’s face recognition.) In this lesson we’ll detect faces in real time with CVZone.

The code

import cv2
from cvzone.FaceDetectionModule import FaceDetector

cap = cv2.VideoCapture(0)
detector = FaceDetector(minDetectionCon=0.5)

while True:
    success, img = cap.read()
    if not success:
        break

    img, faces = detector.findFaces(img)

    if faces:
        face = faces[0]
        print("Centre:", face["center"], "Size:", face["bbox"][2], "x", face["bbox"][3])

    cv2.imshow("Face Detection", img)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

How it works

• FaceDetector(minDetectionCon=0.5) loads MediaPipe’s face model. Faces it’s less than 50% sure about are ignored
• detector.findFaces(img) finds every face, draws a box and a confidence score on the image, and returns the image and a list of faces
• Each face in the list is a dictionary:
  • bbox - the box around the face as (x, y, width, height)
  • center - the middle of the box as (x, y)
  • score - how confident the model is
• Pass draw=False to findFaces() if you want the data without the drawings

Going further

• Multiple faces: faces holds every face in the frame, so len(faces) counts people
• Follow a face: compare face["center"][0] with the middle of the image to tell whether someone is to the left or right. A robot or a pan-tilt camera could turn to follow them
• How close is it? A bigger box means the face is closer to the camera
• Speed things up: a smaller frame is much quicker to process. Add cap.set(cv2.CAP_PROP_FRAME_WIDTH, 640) and cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 480) after opening the camera

Try it yourself

Print “Left”, “Right” or “Centre” depending on where the face is in the picture. The image width is img.shape[1].

< Previous Next >

You can use the arrows  ← → on your keyboard to navigate between lessons.