【tensorflow2.9】搭建第一个卷积网络：手写数字识别

在上一篇已经搭建好了环境，这一偏先来体验一下一个简单的卷积网络。

一、手写数字识别 1.1 导入模块

import numpy as np
from tensorflow import keras
from tensorflow.keras import layers

1.2 设置数据参数

num_classes = 10 
input_shape = (28, 28, 1)

1.3 加载数据并分割

(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train

1.3 图像缩放到 [0, 1]

x_train = x_train.astype("float32") / 255
x_test = x_test.astype("float32") / 255

1.4 确保图像大小

x_train = np.expand_dims(x_train, -1)
x_test = np.expand_dims(x_test, -1)
print("训练集大小:", x_train.shape)
print(x_train.shape[0], "训练样本")
print(x_test.shape[0], "测试样本")

输出：

训练集大小: (60000, 28, 28, 1, 1)
60000 训练样本
10000 测试样本

1.5 类向量转换为二元类矩阵

y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)

1.6 建立模型

model = keras.Sequential(
    [
        keras.Input(shape=input_shape),
        layers.Conv2D(32, kernel_size=(3, 3), activation="relu"),
        layers.MaxPooling2D(pool_size=(2, 2)),
        layers.Conv2D(64, kernel_size=(3, 3), activation="relu"),
        layers.MaxPooling2D(pool_size=(2, 2)),
        layers.Flatten(),
        layers.Dropout(0.5),
        layers.Dense(num_classes, activation="softmax"),
    ]
)

model.summary()

输出： 输出的结果我们暂时不去理解。

1.7 训练模型

batch_size = 128
epochs = 15 # 训练十二轮

model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"]) # 编译

model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, validation_split=0.1) # 训练

1.8 评估模型

以loss和accucry作为指标：

score = model.evaluate(x_test, y_test, verbose=0) 
print("测试 loss:", score[0])
print("测试 accuracy:", score[1])

输出：

测试 loss: 0.02751099318265915
测试 accuracy: 0.9915000200271606