# mnist数字识别
# 设置GPU
import tensorflow as tf
# gpus = tf.config.list_physical_devices('GPU')
# if gpus:
# gpu0 = gpus[0]#如果有多个GPU，仅使用第0个GPU
# tf.config.experimental.set_memory_growth(gpu0,True)#设置GPU显存用量按需使用
# tf.config.set_visible_devices([gpu0],'GPU')
#导入数据 import tensorflow as tf from tensorflow.keras import datasets,layers,models import matplotlib.pyplot as plt (train_images,train_labels), (test_images,test_labels)=datasets.mnist.load_data()
#归一化 #将像素的值标准化至0-1的区间 train_images,test_images = train_images/255.0,test_images/255.0 train_images.shape,test_images.shape,train_labels.shape,test_labels.shape
((60000, 28, 28), (10000, 28, 28), (60000,), (10000,))
#可视化
plt.figure(figsize=(10,10))
for i in range(20):
plt.subplot(5,10,i+1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
plt.imshow(train_images[i],cmap=plt.cm.binary)
plt.xlabel(train_labels[i])
plt.show()
#调整图片格式 train_images = train_images.reshape((60000,28,28,1)) test_images = test_images.reshape((10000,28,28,1))
train_images.shape,test_images.shape,train_labels.shape,test_labels.shape
((60000, 28, 28, 1), (10000, 28, 28, 1), (60000,), (10000,))
#构建CNN网络模型
model = models.Sequential([
layers.Conv2D(32,(3,3),activation='relu',input_shape=(28,28,1)),
layers.MaxPooling2D((2,2)),
layers.Conv2D(64,(3,3),activation='relu'),
layers.MaxPool2D((2,2)),
layers.Flatten(),
layers.Dense(64,activation='relu'),
layers.Dense(10)
])
model.summary()
Model: "sequential" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d (Conv2D) (None, 26, 26, 32) 320 _________________________________________________________________ max_pooling2d (MaxPooling2D) (None, 13, 13, 32) 0 _________________________________________________________________ conv2d_1 (Conv2D) (None, 11, 11, 64) 18496 _________________________________________________________________ max_pooling2d_1 (MaxPooling2 (None, 5, 5, 64) 0 _________________________________________________________________ flatten (Flatten) (None, 1600) 0 _________________________________________________________________ dense (Dense) (None, 64) 102464 _________________________________________________________________ dense_1 (Dense) (None, 10) 650 ================================================================= Total params: 121,930 Trainable params: 121,930 Non-trainable params: 0 _________________________________________________________________ 2022-01-07 23:45:52.268745: I tensorflow/core/platform/cpu_feature_guard.cc:145] This TensorFlow binary is optimized with Intel(R) MKL-DNN to use the following CPU instructions in performance critical operations: SSE4.1 SSE4.2 To enable them in non-MKL-DNN operations, rebuild TensorFlow with the appropriate compiler flags. 2022-01-07 23:45:52.275905: I tensorflow/core/common_runtime/process_util.cc:115] Creating new thread pool with default inter op setting: 8. Tune using inter_op_parallelism_threads for best performance.
#编译模型
model.compile(optimizer='adam',
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy']
)
#训练模型
history = model.fit(train_images,
train_labels,
epochs=10,
validation_data=(test_images,test_labels)
)
Train on 60000 samples, validate on 10000 samples Epoch 1/10 60000/60000 [==============================] - 138s 2ms/sample - loss: 0.1428 - accuracy: 0.9568 - val_loss: 0.0579 - val_accuracy: 0.9809 Epoch 2/10 60000/60000 [==============================] - 134s 2ms/sample - loss: 0.0456 - accuracy: 0.9861 - val_loss: 0.0388 - val_accuracy: 0.9877 Epoch 3/10 60000/60000 [==============================] - 132s 2ms/sample - loss: 0.0328 - accuracy: 0.9897 - val_loss: 0.0304 - val_accuracy: 0.9902 Epoch 4/10 60000/60000 [==============================] - 131s 2ms/sample - loss: 0.0232 - accuracy: 0.9922 - val_loss: 0.0301 - val_accuracy: 0.9897 Epoch 5/10 60000/60000 [==============================] - 134s 2ms/sample - loss: 0.0176 - accuracy: 0.9943 - val_loss: 0.0325 - val_accuracy: 0.9894 Epoch 6/10 60000/60000 [==============================] - 134s 2ms/sample - loss: 0.0135 - accuracy: 0.9956 - val_loss: 0.0307 - val_accuracy: 0.9907 Epoch 7/10 60000/60000 [==============================] - 134s 2ms/sample - loss: 0.0119 - accuracy: 0.9961 - val_loss: 0.0319 - val_accuracy: 0.9905 Epoch 8/10 60000/60000 [==============================] - 134s 2ms/sample - loss: 0.0086 - accuracy: 0.9973 - val_loss: 0.0327 - val_accuracy: 0.9905 Epoch 9/10 60000/60000 [==============================] - 132s 2ms/sample - loss: 0.0080 - accuracy: 0.9974 - val_loss: 0.0296 - val_accuracy: 0.9918 Epoch 10/10 60000/60000 [==============================] - 130s 2ms/sample - loss: 0.0065 - accuracy: 0.9979 - val_loss: 0.0314 - val_accuracy: 0.9911
#预测 plt.imshow(test_images[1])
#输出测试集中第一张图片的预测结果 pre = model.predict(test_images) pre[1]
array([-7.576601 , 0.24191436, 22.48119 , -7.913903 , -6.1873503 ,
-9.024768 , -0.274709 , -1.0671589 , 3.3178928 , -9.065157 ],
dtype=float32)
网络结构
各层的作用
输入层：用于将数据输入到训练网络卷积层：使用卷积核提取图片特征池化层：进行下采样，用更高层的抽象表示图像特征Flatten层：将多维的输入一维化，常用在卷积层到全连接层的过渡全连接层：起到“特征提取器”的作用输出层：输出结果
