Face landmarks detection with dlib

Question

I have the following code:

image_1 = cv2.imread('headshot13-14-2.jpg')
image_1_rgb = cv2.cvtColor(image_1, cv2.COLOR_BGR2RGB)
image_1_gray = cv2.cvtColor(image_1_rgb, cv2.COLOR_BGR2GRAY)
p = "shape_predictor_68_face_landmarks.dat"
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(p)

face = detector(image_1_gray)
face_landmarks = predictor(image_1_gray, face)

And I get the following error for the line face = predictor(image_1_gray, face):

TypeError: __call__(): incompatible function arguments. The following argument types are supported:
    1. (self: dlib.shape_predictor, image: array, box: dlib.rectangle) -> dlib.full_object_detection

However, I checked the type of face (it's dlib.rectangles) and the image_1_gray is a numpy ndarray. Does anyone have any idea why this error still show up?

score 0 · Accepted Answer · answered Dec 10 '20 at 23:44

face variable may contain multiple values, therefore you need to use predictor for each value.

For instance:

for (i, rect) in enumerate(face):
    face_landmarks = predictor(image_1_gray, rect)

To display the detected values on the face:

shp = face_utils.shape_to_np(face_landmarks)

To use face_utils, you need to install imutils.

Most probably your shp variable size is (68, 2). Where 68 is detected points in the face and the 2 is the (x, y) coordinate tuples.

Now, to draw the detected face on the image:

First, get the coordinates

x = rect.left()
y = rect.top()
w = rect.right() - x
h = rect.bottom() - y

Draw the coordinates:

cv2.rectangle(image_1, (x, y), (x + w, y + h), (0, 255, 0), 2)

There may be multiple face in the image, so you may want to label them

cv2.putText(image_1, "Face #{}".format(i + 1), (x - 10, y - 10),
            cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

Now draw the 68 points on the face:

for (x, y) in shp:
    cv2.circle(image_1, (x, y), 1, (0, 0, 255), -1)

Result:

Code:

for (i, rect) in enumerate(face):
    face_landmarks = predictor(image_1_gray, rect)
    shp = face_utils.shape_to_np(face_landmarks)
    x = rect.left()
    y = rect.top()
    w = rect.right() - x
    h = rect.bottom() - y
    cv2.rectangle(image_1, (x, y), (x + w, y + h), (0, 255, 0), 2)
    cv2.putText(image_1, "Face #{}".format(i + 1), (x - 10, y - 10),
                cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
    for (x, y) in shp:
        cv2.circle(image_1, (x, y), 1, (0, 0, 255), -1)

cv2.imshow("face", image_1)
cv2.waitKey(0)

You can also look at the tutorial

Face landmarks detection with dlib

1 Answers1