基于caffe,opencv-python的人脸检测+识别

版权声明：本文为博主原创文章，未经博主允许不得转载。(╰_╯)# https://blog.csdn.net/qq_36735489/article/details/80833989

环境:

python3.5
opencv3.4.1

网络模型:

可以从这个https://www.pyimagesearch.com/2018/02/26/face-detection-with-opencv-and-deep-learning/ 下载,这是一个人脸检测的10层的ResNet+SSD

原理:

利用训练好的caffe的ResNet-10人脸检测网络来检测并抠出人脸
利用opencv的FaceRicognizer()类来对人脸进行识别

读取数据:

def loadData(path):
    dataSet = []
    label = []
    person_list = os.listdir(path)
    for i, a_person in enumerate(person_list):
        dir = os.path.join(path, a_person)
        img_list = os.listdir(dir)
        for a_img in img_list:
            ima = cv2.imread(os.path.join(dir, a_img), 0)
            ima = cv2.resize(ima, (IMAGE_SIZE, IMAGE_SIZE))
            dataSet.append(ima)
            label.append(i)
    label = np.asarray(label)
    return dataSet, label

其中path为所需识别的所有人的总文件夹,子文件夹为所需识别的个人,子文件内存放某人的几张脸部照片(我这里每人只用了五张)

创建网络模型:

net = cv2.dnn.readNetFromCaffe('./models/deploy.prototxt', './models/res10_ssd.caffemodel')

创建->训练->保存人脸识别模型:

# first train
dataSet, label = loadData(r'./img') # 加载人脸图像, 类别
lbfh = cv2.face.LBPHFaceRecognizer_create() # 创建LBPH人脸分类器
lbfh.train(dataSet, label) # 训练模型
lbfh.save('./my_lbph.xml') # 保存模型
# second read and update
# lbfh = cv2.face.LBPHFaceRecognizer_create() # 下次运行时,无需训练之间创建LBPH,然后read即可
# lbfh.read('./my_lbph.xml') # 重载模型
# dataSet, label = updataSet('./3') 
# lbfh.update(dataSet, label) # 还提供了再训练方法
# lbfh.save('./my_lbph.xml')
# lbs = lbfh.getLabels() # 还可以展示所有已训练的图像的各个类别
# print(lbs)

net检测人脸,LBPH识别人脸

  cap = cv2.VideoCapture(0) # 创建摄像头
  result = None
  while 1 :
      _, frame = cap.read()
      frame = cv2.resize(frame, (300, 300))
      (h, w) = frame.shape[:2]
      blob = cv2.dnn.blobFromImage(cv2.resize(frame, (300, 300)), 1.0,  # 送入网络
                                   (300, 300), (104.0, 177.0, 123.0))
      detections = net.forward() # 网络前向预测

      # 或者原图的人脸框
      for i in range(0, detections.shape[2]):
          confidence = detections[0, 0, i, 2]
          # filter out weak detections by ensuring the `confidence` is
          # greater than the minimum confidence
          if confidence < CF: # 置信度阈值
              continue
          box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
          (startX, startY, endX, endY) = box.astype("int")
          testImg = frame[startY:endY, startX:endX] # 将人脸抠出
          if testImg.shape[0]*testImg.shape[1] == 0:
              continue
          #
          testImg = cv2.cvtColor(testImg, cv2.COLOR_BGR2GRAY)
          testImg = cv2.resize(testImg, (IMAGE_SIZE, IMAGE_SIZE))
          # predict face ID
          result = lbfh.predict(testImg) # (id, a_float)
          text = "{:.2f}%-{}".format(confidence * 100, result)
          y = startY - 10 if startY - 10 > 10 else startY + 10
          cv2.rectangle(frame, (startX, startY), (endX, endY),
                        (0, 0, 255), 2)
          cv2.putText(frame, text, (startX, y),
                      cv2.FONT_HERSHEY_SIMPLEX, 0.45, (0, 0, 255), 2)

      # show the output frame
      cv2.imshow("Frame", frame)
      key = cv2.waitKey(1) & 0xFF

      # if the `q` key was pressed, break from the loop
      if key == ord("q"):
          break

总结:

opencv的face recognizer还是很好用的 ,比PCA＋KNN好用,而且简单快捷.