Identifying faces

[geomat0101]

Another performance consideration: 5 vs 68 point facial landmark profiles. Here is the 68:

which breaks down as seen in this method

    if model == 'large':
        return [{
            "chin": points[0:17],
            "left_eyebrow": points[17:22],
            "right_eyebrow": points[22:27],
            "nose_bridge": points[27:31],
            "nose_tip": points[31:36],
            "left_eye": points[36:42],
            "right_eye": points[42:48],
            "top_lip": points[48:55] + [points[64]] + [points[63]] + [points[62]] + [points[61]] + [points[60]],
            "bottom_lip": points[54:60] + [points[48]] + [points[60]] + [points[67]] + [points[66]] + [points[65]] + [points[64]]
        } for points in landmarks_as_tuples]
    elif model == 'small':
        return [{
            "nose_tip": [points[4]],
            "left_eye": points[2:4],
            "right_eye": points[0:2],
        } for points in landmarks_as_tuples]

Each point is an (x,y) coordinate pair, then you compare distances between the point sets generated from two images and come up with a determination of whether or not they are the same.

[geomat0101]

# All the models we need: a detector to find the faces, a shape predictor
# to find face landmarks so we can precisely localize the face, and finally the
# face recognition model.

# these provide face detection and return coordinates of where the faces are with a confidence value
dlib.get_frontal_face_detector()  # "HOG" algorithm - friendlier on CPU, less accurate
dlib.cnn_face_detection_model_v1(cnn_face_detection_model)  # "CNN" algorithm - GPU enabled, more accurate, slower if CPU-only

# these shape predictors find face landmarks so we can precisely localize the face
# 68 is more accurate, and more computationally expensive, than 5
dlib.shape_predictor(predictor_68_point_model)
dlib.shape_predictor(predictor_5_point_model)

# the face recognition model
# the object returned has a method compute_face_descriptor that takes the output of the shape_predictor
# and computes a 128D vector that describes the face in img identified by the shape.
dlib.face_recognition_model_v1(face_recognition_model)