面部特征点检测dlib库使用

面部特征点检测dlib库使用

1、运行环境

python3
dlib
cv2
如果没有安装dlib和cv2, 建议使用conda 安装，简单不容易报错。

conda install -c conda-forge dlib
conda install -c conda-forge opencv

2、面部特征点检测

def landmark_detector(img_path):
    predictor_path = './68_landmarks/shape_predictor_68_face_landmarks.dat'
    detector = dlib.get_frontal_face_detector()   # 加载面部检测器
    predictor = dlib.shape_predictor(predictor_path) # 加载特征点检测器
    img = cv2.imread(img_path)
    dets = detector(img, 1)                       # 检测人脸，肯能有多个
    landmarks = []
    for k, d in enumerate(dets):
        shape = predictor(img, d)
        tmp = []
	    for point in shape.parts():
	        x, y = point.x, point.y
	        tmp.append((x, y))
	    landmarks.append(tmp)
    return landmarks

predictor download path:shape_predictor_68_face_landmarks