基于jupyter-notebook的口罩数据集的划分、训练、测试，并实现摄像头口罩识别判断_jupyter notebook如何分离数据集

一、笑脸数据集（genki4k）正负样本的划分、模型训练和测试的过程（至少包括SVM、CNN），输出模型训练精度和测试精度（F1-score和ROC）
1.数据集下载

笑脸数据集下载链接：https://pan.baidu.com/s/1ldA8CF6pH4q0Bheik_Tttg
提取码：fui1
2.训练数据集

1. 导入Keras

import keras
keras.__version__
1
2

2) 读取数据集，训练

import os, shutil
# The path to the directory where the original
# dataset was uncompressed
original_dataset_dir = 'smileface/genki4k'

# The directory where we will
# store our smaller dataset
base_dir = 'smileface/smile_and_nosmile'
os.mkdir(base_dir)

# Directories for our training,
# validation and test splits
train_dir = os.path.join(base_dir, 'train')
os.mkdir(train_dir)
validation_dir = os.path.join(base_dir, 'validation')
os.mkdir(validation_dir)
test_dir = os.path.join(base_dir, 'test')
os.mkdir(test_dir)

# Directory with our training smile pictures
train_smile_dir = os.path.join(train_dir, 'smiles')
os.mkdir(train_smile_dir)

# Directory with our training nosmile pictures
train_nosmile_dir = os.path.join(train_dir, 'no_smiles')
os.mkdir(train_nosmile_dir)

# Directory with our validation smile pictures
validation_smile_dir = os.path.join(validation_dir, 'smiles')
os.mkdir(validation_smile_dir)

# Directory with our validation nosmile pictures
validation_nosmile_dir = os.path.join(validation_dir, 'no_smiles')
os.mkdir(validation_nosmile_dir)

# Directory with our validation smile pictures
test_smile_dir = os.path.join(test_dir, 'smiles')
os.mkdir(test_smile_dir)

# Directory with our validation nosmile pictures
test_nosmile_dir = os.path.join(test_dir, 'no_smiles')
os.mkdir(test_nosmile_dir)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

3. 将笑脸图片和非笑脸图片放入对应文件夹

在上面程序中生成了一个名为smile_and_nosmile的文件夹，里面有三个子文件，分别存放训练、测试、验证数据，在这三个文件夹下还有smile和nosmile文件夹，我们需要将笑脸图片放入smile文件夹，将非笑脸图片放入nosmile文件夹。

4) 打印数据集中笑脸和非笑脸图片数

print('total training smile images:', len(os.listdir(train_smile_dir)))
print('total training nosmile images:', len(os.listdir(train_nosmile_dir)))
print('total validation smile images:', len(os.listdir(validation_smile_dir)))
print('total validation nosmile images:', len(os.listdir(validation_nosmile_dir)))
print('total test smile images:', len(os.listdir(test_smile_dir)))
print('total test nosmile images:', len(os.listdir(test_nosmile_dir)))
1
2
3
4
5
6

选用了几十张图像进行模型的构建，自然其最终精度也会下降许多
5) 构建小型卷积神经网络

from keras import layers
from keras import models

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu',
                        input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

了解征图的尺寸是如何随着每一层变化的

model.summary()
1

from keras import optimizers
model.compile(loss='binary_crossentropy',
              optimizer=optimizers.RMSprop(lr=1e-4),
              metrics=['acc'])
1
2
3
4

6. 数据预处理

from keras.preprocessing.image import ImageDataGenerator
# All images will be rescaled by 1./255
train_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)

train_generator = train_datagen.flow_from_directory(
        # This is the target directory
        train_dir,
        # All images will be resized to 150x150
        target_size=(150, 150),
        batch_size=20,
        # Since we use binary_crossentropy loss, we need binary labels
        class_mode='binary')

validation_generator = test_datagen.flow_from_directory(
        validation_dir,
        target_size=(150, 150),
        batch_size=20,
        class_mode='binary')
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

生成器的输出

for data_batch, labels_batch in train_generator:
    print('data batch shape:', data_batch.shape)
    print('labels batch shape:', labels_batch.shape)
    break
1
2
3
4

7) 训练模型

history = model.fit_generator(
      train_generator,
      steps_per_epoch=100,
      epochs=30,
      validation_data=validation_generator,
      validation_steps=50)
1
2
3
4
5
6

这里使用fit_generator方法来完成此操作，对于我们这样的数据生成器，它相当于fit方法。

保存模型

model.save('smile_and_nosmile.h5')
1

保存好模型之后便可以进行此模型的人脸识别了，不过精度不会太高，毕竟只使用了这么少量的图片

绘制模型的损失和准确性

import matplotlib.pyplot as plt

acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(len(acc))

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()

plt.figure()

plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
plt.show()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

8) 数据增强

数据增强采用的方法是从现有的训练样本中生成更多的训练数据，方法是通过一系列随机变换来“增强”样本，从而产生看上去可信的图像。我们的目标是在训练时，我们的模型不会两次看到完全相同的图像。这有助于将模型暴露于数据的更多方面，并更好地泛化。

datagen = ImageDataGenerator(
      rotation_range=40,
      width_shift_range=0.2,
      height_shift_range=0.2,
      shear_range=0.2,
      zoom_range=0.2,
      horizontal_flip=True,
      fill_mode='nearest')
1
2
3
4
5
6
7
8

查看数据增强后的图像

# This is module with image preprocessing utilities
from keras.preprocessing import image

fnames = [os.path.join(train_smile_dir, fname) for fname in os.listdir(train_smile_dir)]

# We pick one image to "augment"
img_path = fnames[3]

# Read the image and resize it
img = image.load_img(img_path, target_size=(150, 150))

# Convert it to a Numpy array with shape (150, 150, 3)
x = image.img_to_array(img)

# Reshape it to (1, 150, 150, 3)
x = x.reshape((1,) + x.shape)

# The .flow() command below generates batches of randomly transformed images.
# It will loop indefinitely, so we need to `break` the loop at some point!
i = 0
for batch in datagen.flow(x, batch_size=1):
    plt.figure(i)
    imgplot = plt.imshow(image.array_to_img(batch[0]))
    i += 1
    if i % 4 == 0:
        break

plt.show()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

如果我们使用这种数据增加配置训练一个新的网络，我们的网络将永远不会看到两次相同的输入。然而，它看到的输入仍然是高度相关的，因为它们来自少量的原始图像——我们不能产生新的信息，我们只能混合现有的信息。因此，这可能还不足以完全消除过度拟合。

为了进一步对抗过拟合，我们还将在我们的模型中增加一个Dropout层，就在密集连接分类器之前：

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu',
                        input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

model.compile(loss='binary_crossentropy',
              optimizer=optimizers.RMSprop(lr=1e-4),
              metrics=['acc'])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

用数据增强和退出训练网络

train_datagen = ImageDataGenerator(
    rescale=1./255,
    rotation_range=40,
    width_shift_range=0.2,
    height_shift_range=0.2,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True,)

# Note that the validation data should not be augmented!
test_datagen = ImageDataGenerator(rescale=1./255)

train_generator = train_datagen.flow_from_directory(
        # This is the target directory
        train_dir,
        # All images will be resized to 150x150
        target_size=(150, 150),
        batch_size=32,
        # Since we use binary_crossentropy loss, we need binary labels
        class_mode='binary')

validation_generator = test_datagen.flow_from_directory(
        validation_dir,
        target_size=(150, 150),
        batch_size=32,
        class_mode='binary')

history = model.fit_generator(
      train_generator,
      steps_per_epoch=100,
      epochs=100,
      validation_data=validation_generator,
      validation_steps=50)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33

保存模型

model.save('smile_and_nosmile_1.h5')
1

再查看模型的损失和准确性

acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(len(acc))

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()

plt.figure()

plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
plt.show()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

3.摄像头实时笑脸识别

import cv2
from keras.preprocessing import image
from keras.models import load_model
import numpy as np
import dlib
from PIL import Image
model = load_model('smile_and_nosmile_1.h5')
detector = dlib.get_frontal_face_detector()
video=cv2.VideoCapture(0)
font = cv2.FONT_HERSHEY_SIMPLEX
def rec(img):
    gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
    dets=detector(gray,1)
    if dets is not None:
        for face in dets:
            left=face.left()
            top=face.top()
            right=face.right()
            bottom=face.bottom()
            cv2.rectangle(img,(left,top),(right,bottom),(0,255,0),2)
            img1=cv2.resize(img[top:bottom,left:right],dsize=(150,150))
            img1=cv2.cvtColor(img1,cv2.COLOR_BGR2RGB)
            img1 = np.array(img1)/255.
            img_tensor = img1.reshape(-1,150,150,3)
            prediction =model.predict(img_tensor)    
            print(prediction)
            if prediction[0][0]<0.5:
                result='no_smiles'
            else:
                result='smiles'
            cv2.putText(img, result, (left,top), font, 2, (0, 255, 0), 2, cv2.LINE_AA)
        cv2.imshow('smile detector', img)
while video.isOpened():
    res, img_rd = video.read()
    if not res:
        break
    rec(img_rd)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
video.release()
cv2.destroyAllWindows()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41

二、将数据集换为口罩数据集，并进行训练，实现佩戴口罩的人脸识别
1.口罩数据集下载

人脸口罩数据集下载链接：https://pan.baidu.com/s/11PBCmDDx7Dtx_ckjwZR2uw
提取码：n2um

训练人脸口罩数据集和训练笑脸数据集一样，只需改一下和相应的变量名和数据集。
2.口罩佩戴的人脸识别

import cv2
from keras.preprocessing import image
from keras.models import load_model
import numpy as np
import dlib
from PIL import Image
model = load_model('mask_and_nomask.h5')
detector = dlib.get_frontal_face_detector()
video=cv2.VideoCapture(0)
font = cv2.FONT_HERSHEY_SIMPLEX
def rec(img):
    gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
    dets=detector(gray,1)
    if dets is not None:
        for face in dets:
            left=face.left()
            top=face.top()
            right=face.right()
            bottom=face.bottom()
            cv2.rectangle(img,(left,top),(right,bottom),(0,255,0),2)
            img1=cv2.resize(img[top:bottom,left:right],dsize=(150,150))
            img1=cv2.cvtColor(img1,cv2.COLOR_BGR2RGB)
            img1 = np.array(img1)/255.
            img_tensor = img1.reshape(-1,150,150,3)
            prediction =model.predict(img_tensor)    
            print(prediction)
            if prediction[0][0]>0.5:
                result='nomask'
            else:
                result='mask'
            cv2.putText(img, result, (left,top), font, 2, (0, 255, 0), 2, cv2.LINE_AA)
        cv2.imshow('mask detector', img)
while video.isOpened():
    res, img_rd = video.read()
    if not res:
        break
    rec(img_rd)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
video.release()
cv2.destroyAllWindows()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41

参考文献

https://blog.csdn.net/dQCFKyQDXYm3F8rB0/article/details/104743982
https://blog.csdn.net/cj151525/article/details/104984897
https://blog.csdn.net/weixin_45137708/article/details/107142706
https://blog.csdn.net/weixin_45306136/article/details/107170815