15 手写数字识别

1.手写数字数据集

from tensorflow.keras.datasets import mnist
(X_tarin, y_train), (X_test, y_test) = mnist.load_data()

2.图片数据预处理

• x：归一化MinMaxScaler()
• y：独热编码OneHotEncoder()或to_categorical
• 训练集测试集划分
• 张量结构

from sklearn.preprocessing import MinMaxScaler, OneHotEncoder
import numpy as np
scaler = MinMaxScaler()
# 将数组重新整形为2d所需的三维数组
nsamples1, nx1, ny1 = X_tarin.shape
X_tarin = X_tarin.reshape((nsamples1,nx1*ny1))
nsamples2, nx2, ny2 = X_test.shape
X_test = X_test.reshape((nsamples2,nx2*ny2))

X_tarin = scaler.fit_transform(X_tarin)
X_test = scaler.fit_transform(X_test)
print("训练集归一化后",X_tarin)
print("测试集归一化后",X_test)

X_tarin=X_tarin.reshape(-1,28,28,1)
X_test=X_test.reshape(-1,28,28,1)

y_train = y_train.astype(np.float32).reshape(-1,1)  #将y_train变为一列
y_test = y_test.astype(np.float32).reshape(-1,1)  #将y_test变为一列
y_train = OneHotEncoder().fit_transform(y_train).todense() #张量结构todense
y_test = OneHotEncoder().fit_transform(y_test).todense() #张量结构todense

print("独热编码：",y_train)
print("独热编码：",y_test)

　　

       

3.设计卷积神经网络结构

• 绘制模型结构图，并说明设计依据。

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense,Dropout,Conv2D,MaxPool2D,Flatten
#3、建立模型
model = Sequential()
ks = (3, 3)  # 卷积核的大小
# input_shape = X_tarin.shape[1:]
# 一层卷积，padding='same',tensorflow会对输入自动补0
model.add(Conv2D(filters=16, kernel_size=ks, padding='same', input_shape=(28, 28, 1), activation='relu'))
# 池化层1
model.add(MaxPool2D(pool_size=(2, 2)))
# 防止过拟合，随机丢掉连接
model.add(Dropout(0.25))
# 二层卷积
model.add(Conv2D(filters=32, kernel_size=ks, padding='same', activation='relu'))
# 池化层2
model.add(MaxPool2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# 三层卷积
model.add(Conv2D(filters=64, kernel_size=ks, padding='same', activation='relu'))
# 四层卷积
model.add(Conv2D(filters=128, kernel_size=ks, padding='same', activation='relu'))
# 池化层3
model.add(MaxPool2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# 平坦层
model.add(Flatten())
# 全连接层
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.25))
# 激活函数softmax
model.add(Dense(10, activation='softmax'))
# 输出模型各层的参数状况
print(model.summary())

4.模型训练

• model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
• train_history = model.fit(x=X_train,y=y_train,validation_split=0.2, batch_size=300,epochs=10,verbose=2)

# 4、模型训练
import tensorflow as tf
check_path = 'ckpt/cp-{epoch:04d}.ckpt'
# period 每隔5epoch保存一次
save_model_cb = tf.keras.callbacks.ModelCheckpoint(check_path, save_weights_only=True, verbose=1, period=5)

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
train_history = model.fit(x=X_tarin, y=y_train, validation_split=0.2, batch_size=300, epochs=10, verbose=2,callbacks=[save_model_cb])
# 准确率
show_train_history(train_history, 'accuracy', 'val_accuracy')
# 损失率
show_train_history(train_history, 'loss', 'val_loss')

5.模型评价

• model.evaluate()
• 交叉表与交叉矩阵
• pandas.crosstab
• seaborn.heatmap

# 5、模型评价
import pandas as pd
import seaborn as sns
# model.evaluate()
score = model.evaluate(X_test, y_test)
print('score：', score)
# 预测值
y_pred = model.predict_classes(X_test)
print('y_pred：', y_pred[:10])
# 交叉表与交叉矩阵
y_test1 = np.argmax(y_test, axis=1).reshape(-1)
y_true = np.array(y_test1)[0]
# 交叉表查看预测数据与原数据对比
# pandas.crosstab
pd.crosstab(y_true, y_pred, rownames=['true'], colnames=['predict'])
# 交叉矩阵
# seaborn.heatmap
y_test1 = y_test1.tolist()[0]
a = pd.crosstab(np.array(y_test1), y_pred, rownames=['Lables'], colnames=['Predict'])
# 转换成属dataframe
df = pd.DataFrame(a)
sns.heatmap(df, annot=True, cmap="Reds", linewidths=0.2, linecolor='G')
plt.show()