def build_camera(self):
# The location of the face cascade file in the opencv file, used to help identify faces in images or video streams
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades +
'haarcascade_frontalface_alt.xml')
# Read the subfolder name under the dataset
name_list = read_name_list('./dataset')
# Turn on the camera and start reading
cameraCapture = cv2.VideoCapture(700)
success, frame = cameraCapture.read()
while success and cv2.waitKey(1) == -1:
success, frame = cameraCapture.read()
# Grayscale image
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# Face detection: scaleFactor is
#the magnification ratio (this parameter must be greater than 1);
#minNeighbors is the number of repeated identifications (this parameter is used to adjust the accuracy, the larger the more accurate)
faces = face_cascade.detectMultiScale(gray,
scaleFactor=1.3,
minNeighbors=8)
for (x, y, w, h) in faces:
origin = gray[x:x + w, y:y + h]
# origin is the original image, self.img_size is the size of the output image, and interpolation is the interpolation method
origin = cv2.resize(origin, (self.img_size, self.img_size),
interpolation=cv2.INTER_LINEAR)
# Use the model to compare the faces recognized by cv2
label, prob = self.model.predict(origin)
# If the model believes that the probability is higher than 70%, it will be displayed as the existing label in the model
if prob > 0.7:
show_name = name_list[label]
else:
show_name = 'unknown'
# Frame a face
frame = cv2.rectangle(frame, (x, y), (x + w, y + h),
(255, 0, 0), 1)
if prob > 0.7:
#storing recognized folder in folder named with date of that day
today = datetime.now()
recpath = './Recognized/' + today.strftime("%d-%m-%Y")
if not os.path.exists(recpath):
os.makedirs(recpath)
#print(os.getcwd())
cv2.imwrite(
recpath + '/' + today.strftime("%H%M%S") + '.jpg',
frame[y:y + h, x:x + w])
#if not os.path.exists()
# Show name
cv2.putText(frame, show_name, (x, y - 20),
cv2.FONT_HERSHEY_SIMPLEX, 1, 255, 2)
cv2.imshow("Recognizing...", frame)
# Release camera
cameraCapture.release()
# Close all windows
cv2.destroyAllWindows()
def test_onePicture(path):
# 加载模型
model = Model()
model.load()
# 读取图片
img = cv2.imread(path)
# 重置图片尺寸为：128*128
img = cv2.resize(img, (128, 128))
# 图片灰度化
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# labelIndex为概率最高的label的索引号, prob为对应概率
labelIndex, prob = model.predict(img)
if labelIndex != -1:
name_list = read_name_list('./dataset')
print(name_list[labelIndex], prob)
else:
print("Don't know this person.")
def test_onBatch(path):
# 加载模型
model = Model()
model.load()
# 计数器
index = 0
# 读取所有图片；img_list是所有图片的集合，label_lsit是所有标签的集合，label_num是标签数量
img_list, label_lsit, label_num = read_file(path)
for img in img_list:
# labelIndex为概率最高的label的索引号, prob为对应概率
labelIndex, prob = model.predict(img)
if labelIndex != -1:
index += 1
name_list = read_name_list('./dataset')
print(name_list[labelIndex])
else:
print("Don't know this person.")
return index
def build_camera(self):
# opencv文件中人脸级联文件的位置，用于帮助识别图像或者视频流中的人脸
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades +
'haarcascade_frontalface_alt.xml')
# 读取dataset数据集下的子文件夹名称，即标签名
name_list = read_name_list('./dataset')
# 打开摄像头并开始读取画面
cameraCapture = cv2.VideoCapture('recognize_faces2.mp4')
success, frame = cameraCapture.read()
while success and cv2.waitKey(1) == -1:
success, frame = cameraCapture.read()
# 图像灰度化
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# 检测人脸：scaleFactor是放大比例（此参数必须大于1）；minNeighbors是重复识别次数（此参数用来调整精确度，越大则越精确）
faces = face_cascade.detectMultiScale(gray,
scaleFactor=1.3,
minNeighbors=8)
for (x, y, w, h) in faces:
origin = gray[x:x + w, y:y + h]
# origin是原图，self.img_size是输出图像的尺寸，interpolation是插值方法
origin = cv2.resize(origin, (self.img_size, self.img_size),
interpolation=cv2.INTER_LINEAR)
# 利用模型对cv2识别出的人脸进行比对
label, prob = self.model.predict(origin)
# 如果模型认为概率高于70%则显示为模型中已有的label
if prob > 0.7:
show_name = name_list[label]
else:
show_name = 'unknown'
# 框出人脸
frame = cv2.rectangle(frame, (x, y), (x + w, y + h),
(255, 0, 0), 1)
# 显示人名
cv2.putText(frame, show_name, (x, y - 20),
cv2.FONT_HERSHEY_SIMPLEX, 1, 255, 2)
cv2.imshow("Recognizing...", frame)
# 释放摄像头
cameraCapture.release()
# 关闭所有窗口
cv2.destroyAllWindows()
model.load()
# 读取图片
image = cv2.imread(path)
# 重置图片尺寸为：128*128、512*512
small_image = cv2.resize(image, (128, 128))
# 图片灰度化
small_gray = cv2.cvtColor(small_image, cv2.COLOR_BGR2GRAY)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# 检测人脸，探测图片中的人脸
faces = face_engine.detectMultiScale(gray, scaleFactor=1.3, minNeighbors=15)
# 读取dataset数据集下的子文件夹名称
name_list = read_name_list('./dataset')
# 框出人脸
for (x, y, w, h) in faces:
origin = gray[x:x + w, y:y + h]
# origin是原图，self.img_size是输出图像的尺寸，interpolation是插值方法
origin = cv2.resize(origin, (128, 128), interpolation=cv2.INTER_LINEAR)
# 利用模型对cv2识别出的人脸进行比对
# 第一个返回值为概率最高的label的index,第二个返回值为对应概率
labelIndex, prob = model.predict(origin)
# 如果模型认为概率高于70%则显示为模型中已有的label
if prob > 0.7:
show_name = name_list[labelIndex]
else:
show_name = 'unknown'
# 1.原始图片 2.人脸坐标原点 3.标记的高度 4，线的颜色 5，线宽
frame = cv2.rectangle(image, (x, y), (x + w, y + h), (0, 255, 0), 1)