Python人脸相似度对比

# -*- coding: utf-8 -*- import os,dlib,glob,numpy from skimage import io # 人脸关键点检测器 predictor_path = "shape_predictor.dat" # 人脸识别模型、提取特征值 face_rec_model_path = "dlib_face_recognition.dat" # 训练图像文件夹 faces_folder_path ='train_images' # 加载模型 detector = dlib.get_frontal_face_detector() sp = dlib.shape_predictor(predictor_path) facerec = dlib.face_recognition_model_v1(face_rec_model_path) candidate = [] # 存放训练集人物名字 descriptors = [] #存放训练集人物特征列表 for f in glob.glob(os.path.join(faces_folder_path,"*.jpg")): print("正在处理: {}".format(f)) img = io.imread(f) candidate.append(f.split('\\')[-1].split('.')[0]) # 人脸检测 dets = detector(img, 1) for k, d in enumerate(dets): shape = sp(img, d) # 提取特征 face_descriptor = facerec.compute_face_descriptor(img, shape) v = numpy.array(face_descriptor) descriptors.append(v) print('识别训练完毕！') try: ## test_path=input('请输入要检测的图片的路径（记得加后缀哦）:') img = io.imread(r".\test_images\test6.jpg") dets = detector(img, 1) except: print('输入路径有误，请检查！') dist = [] for k, d in enumerate(dets): shape = sp(img, d) face_descriptor = facerec.compute_face_descriptor(img, shape) d_test = numpy.array(face_descriptor) for i in descriptors: #计算距离 dist_ = numpy.linalg.norm(i-d_test) dist.append(dist_) # 训练集人物和距离组成一个字典 c_d = dict(zip(candidate,dist)) cd_sorted = sorted(c_d.items(), key=lambda d:d[1]) print ("识别到的人物最有可能是: ",cd_sorted[0][0]) print(c_d)