【tensorflow】T1.实现mnist手写数字识别

>- **🍨 本文为[🔗365天深度学习训练营](https://mp.weixin.qq.com/s/0dvHCaOoFnW8SCp3JpzKxg) 中的学习记录博客**
>- **🍖 原作者：[K同学啊](https://mtyjkh.blog.csdn.net/)**

我的环境：

语言环境：python 3.11.5

编译器：Spyder

深度学习环境：Tensorflow 2.12.0

1. 前期工作

导入数据：

import tensorflow as tf
from tensorflow.keras import datasets, models, layers

(train_images, train_labels), (test_images, test_labels) = datasets.mnist.load_data()

归一化：

train_images, test_images = train_images / 255.0, test_images / 255.0

train_images.shape,
test_images.shape,
train_labels.shape,
test_labels.shape

    Out[7]: ((60000, 28, 28, 1), (10000, 28, 28, 1), (60000,), (10000,))

可视化：

import matplotlib.pyplot as plt

plt.figure(figsize=(20, 10))

for i in range(20):
    
    plt.subplot(5, 10, i+1)
    plt.imshow(train_images[i], cmap = plt.cm.binary)
    plt.xlabel(train_labels[i])
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    
plt.show()

调整格式：

train_images = train_images.reshape((60000, 28, 28, 1))
test_images = test_images.reshape((10000, 28, 28, 1))

2. 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.MaxPooling2D((2, 2)),
    
    layers.Flatten(),
    layers.Dense(64, activation = "relu"),
    layers.Dense(10)
    
    ])

model.summary()

Model: "sequential_2"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d_4 (Conv2D)           (None, 26, 26, 32)        320       
                                                                 
 max_pooling2d_4 (MaxPooling  (None, 13, 13, 32)       0         
 2D)                                                             
                                                                 
 conv2d_5 (Conv2D)           (None, 11, 11, 64)        18496     
                                                                 
 max_pooling2d_5 (MaxPooling  (None, 5, 5, 64)         0         
 2D)                                                             
                                                                 
 flatten_2 (Flatten)         (None, 1600)              0         
                                                                 
 dense_4 (Dense)             (None, 64)                102464    
                                                                 
 dense_5 (Dense)             (None, 10)                650       
                                                                 
=================================================================
Total params: 121,930
Trainable params: 121,930
Non-trainable params: 0
_________________________________________________________________

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)
    
    )

训练结果：

2024-05-09 14:26:41.574956: W tensorflow/tsl/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz
Epoch 1/10
1875/1875 [==============================] - 14s 7ms/step - loss: 0.1369 - accuracy: 0.9584 - val_loss: 0.0387 - val_accuracy: 0.9884
Epoch 2/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0448 - accuracy: 0.9861 - val_loss: 0.0414 - val_accuracy: 0.9874
Epoch 3/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0321 - accuracy: 0.9901 - val_loss: 0.0394 - val_accuracy: 0.9874
Epoch 4/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0251 - accuracy: 0.9922 - val_loss: 0.0349 - val_accuracy: 0.9888
Epoch 5/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0185 - accuracy: 0.9942 - val_loss: 0.0295 - val_accuracy: 0.9912
Epoch 6/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0140 - accuracy: 0.9955 - val_loss: 0.0347 - val_accuracy: 0.9900
Epoch 7/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0115 - accuracy: 0.9960 - val_loss: 0.0329 - val_accuracy: 0.9903
Epoch 8/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0098 - accuracy: 0.9968 - val_loss: 0.0337 - val_accuracy: 0.9909
Epoch 9/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0078 - accuracy: 0.9973 - val_loss: 0.0359 - val_accuracy: 0.9905
Epoch 10/10
1875/1875 [==============================] - 13s 7ms/step - loss: 0.0080 - accuracy: 0.9971 - val_loss: 0.0339 - val_accuracy: 0.9911

预测：

plt.imshow(test_images[7])

pre = model.predict(
    test_images,
    
    )

pre[7]

313/313 [==============================] - 1s 3ms/step

（图片是9...懒得传了hhhh）

Out[12]: 
array([ -9.191909 ,  -1.6822042,   2.1744351,   1.866617 ,   3.8742313,
         1.7728972, -11.349705 ,  -2.4591968,   1.5152467,   8.1270685],
      dtype=float32)

总结：

其实前面的概念大致都清楚，就是在最后的array里才算真的明白了cnn怎么输出test结果，也算接触dl之后通过实践解决的第一个疑惑。