Deep.Learning.for.Computer.Vision.with.Python-----chapter14 Lenet网络模型

一、代码演示

定义网络 lenet.py：

"""
LeNet:INPUT => CONV => ReLU => POOL => CONV => ReLU => POOL =>FC => ReLU => FC
INPUT:28X28X1
CONV:28X28X20  Filter:5X5 K=20
ACT:ReLU
POOL:14X14X20  2X2 步长是2X2 每行移动2个 换行后也是2个
CONV:14X14X50 5X5 K=50
ACT:ReLU
POOL:7X7X50  2X2
FC:500
ACT:ReLU
FC:10
SOFTMAX:10
"""
from keras.models import Sequential
from keras.layers.convolutional import Conv2D
from keras.layers.convolutional import MaxPooling2D
from keras.layers.core import Activation
from keras.layers.core import Flatten
from keras.layers.core import Dense
from keras import backend as K

#构建网络
class LeNet:
    @staticmethod
    #class:标签类别数量
    def build(width,height,depth,classes):
        model = Sequential()
        inputShape = (height,width,depth)
        if K.image_data_format()=='channels_first':
            inputShape = (depth,height,width)
        #CONV => RELU => POOL
        model.add(Conv2D(20,(5,5),padding='same',input_shape=inputShape))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2)))
        # CONV => RELU => POOL
        model.add(Conv2D(50,(5,5),padding='same'))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2)))
        #输入量变平，并可以应用一个具有500个节点的全连接层
        #FC => RELU
        model.add(Flatten())
        model.add(Dense(500))
        model.add(Activation('relu'))
        #=> FC classes=10(10类标签）
        model.add(Dense(classes))
        model.add(Activation('softmax'))
        return model

训练网络 lenet_mnist.py：

"""
1.加载mnist数据
2.实例化网络
3.训练网络
4.评估
"""
from lenet import LeNet
from keras.optimizers import SGD
from sklearn.preprocessing import LabelBinarizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
import scipy.io as scio
from keras import backend as K
import matplotlib.pyplot as plt
import numpy as np

#1.加载数据
print('[INFO] accessing MNIST...')
dataset = scio.loadmat('D:/DLCV/DLCVmy_study/datasets/openml/mnist-original.mat')
data = dataset['data'].T
label = dataset['label'].T
#print(data.shape)
if K.image_data_format()=='channels_first':
    #data.shape[0]图片个数
    data = data.reshape(data.shape[0],1,28,28)
else:
    data = data.reshape(data.shape[0], 28, 28, 1)
"将原始图像强度压缩到0-1"
data = data.astype("float")/255.0
(trainX, testX, trainY, testY) = train_test_split(data, label, test_size=0.25,random_state=42)

#将标签转为向量 如:若标签类别是3 则对应输出[0, 0, 0, 1, 0, 0, 0, 0, 0, 0]
le = LabelBinarizer()
trainY = le.fit_transform(trainY)
testY = le.transform(testY)

#2.初始化优化器和网络
print('[INFO] compiling model...')
opt = SGD(lr=0.01)
model = LeNet.build(width=28,height=28,depth=1,classes=10)
model.compile(loss='categorical_crossentropy',optimizer=opt,metrics=['accuracy'])
#3.x训练网络
print('[INFO] training network...')
# verbose:1代表显示进度条，0不显示进度条，默认为0
H = model.fit(trainX,trainY,validation_data=(testX,testY),batch_size=128,epochs=20,verbose=1)
#4.评估网络
print('[INFO] evaluating network...')
predictions = model.predict(testX,batch_size=128)
print(classification_report(testY.argmax(axis=1),predictions.argmax(axis=1),target_names=[str(x) for x in le.classes_]))

plt.style.use('ggplot')
plt.figure()
plt.plot(np.arange(0,20),H.history['loss'],label='train_loss')
plt.plot(np.arange(0,20),H.history['val_loss'],label='val_loss')
plt.plot(np.arange(0,20),H.history['acc'],label='train_acc')
plt.plot(np.arange(0,20),H.history['val_acc'],label='val_acc')
plt.title("Training Loss and Accuracy")
plt.xlabel('Epoch #')
plt.ylabel('Loss/Accuracy')
plt.legend()
plt.show()