卷积神经网络，TensorFlow实现

「已注销」

已于 2022-09-20 10:07:01 修改

阅读量382

点赞数 1

分类专栏：深度学习文章标签： tensorflow 深度学习 python

于 2022-09-20 09:59:29 首次发布

本文链接：https://blog.youkuaiyun.com/m0_73817310/article/details/126947102

版权

深度学习专栏收录该内容

24 篇文章

订阅专栏

本文介绍了如何利用TensorFlow下载和准备CIFAR10数据集，并构建卷积神经网络模型进行深度学习训练。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

import tensorflow as tf
import matplotlib.pyplot as plt
from tensorflow import keras
from tensorflow.keras import datasets
from tensorflow.keras import layers
print("TensorFlow version:", tf.__version__)

TensorFlow version: 2.6.4

下载并准备 CIFAR10 数据集

CIFAR10 数据集包含 10 类，共 60000 张彩色图片，每类图片有 6000 张。此数据集中 50000 个样例被作为训练集，剩余 10000 个样例作为测试集。类之间相互独立，不存在重叠的部分。

train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()

# Normalize pixel values to be between 0 and 1
train_images, test_images = train_images / 255.0, test_images / 255.0

Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
170500096/170498071 [==============================] - 6s 0us/step
170508288/170498071 [==============================] - 6s 0us/step

构造卷积神经网络模型

inputs = keras.Input(shape=(32, 32, 3))
x = layers.Conv2D(filters=32, kernel_size=3, activation="relu")(inputs)
x = layers.MaxPooling2D(pool_size=2)(x)
x = layers.Conv2D(filters=64, kernel_size=3, activation="relu")(x)
x = layers.MaxPooling2D(pool_size=2)(x)
x = layers.Conv2D(filters=128, kernel_size=3, activation="relu")(x)
x = layers.Flatten()(x)
x = layers.Dropout(0.5)(x)
outputs = layers.Dense(10, activation="softmax")(x)
model = keras.Model(inputs=inputs, outputs=outputs)
model.summary()

Model: "model"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 32, 32, 3)]       0         
_________________________________________________________________
conv2d (Conv2D)              (None, 30, 30, 32)        896       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 13, 13, 64)        18496     
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 4, 4, 128)         73856     
_________________________________________________________________
flatten (Flatten)            (None, 2048)              0         
_________________________________________________________________
dropout (Dropout)            (None, 2048)              0         
_________________________________________________________________
dense (Dense)                (None, 10)                20490     
=================================================================
Total params: 113,738
Trainable params: 113,738
Non-trainable params: 0

编译并训练模型

model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(),
              metrics=['accuracy'])

history = model.fit(train_images, train_labels, epochs=30, 
                    validation_split=0.3,  batch_size = 128)

Epoch 1/30
274/274 [==============================] - 3s 8ms/step - loss: 1.8420 - accuracy: 0.3212 - val_loss: 1.5064 - val_accuracy: 0.4585
Epoch 2/30
274/274 [==============================] - 1s 5ms/step - loss: 1.4373 - accuracy: 0.4795 - val_loss: 1.3612 - val_accuracy: 0.5140
Epoch 3/30
274/274 [==============================] - 1s 5ms/step - loss: 1.2847 - accuracy: 0.5413 - val_loss: 1.2257 - val_accuracy: 0.5707
Epoch 4/30
274/274 [==============================] - 1s 5ms/step - loss: 1.1985 - accuracy: 0.5726 - val_loss: 1.1569 - val_accuracy: 0.5939
Epoch 5/30
274/274 [==============================] - 1s 5ms/step - loss: 1.1186 - accuracy: 0.6062 - val_loss: 1.0632 - val_accuracy: 0.6331

...

Epoch 25/30
274/274 [==============================] - 1s 5ms/step - loss: 0.6362 - accuracy: 0.7758 - val_loss: 0.7557 - val_accuracy: 0.7424
Epoch 26/30
274/274 [==============================] - 1s 5ms/step - loss: 0.6278 - accuracy: 0.7801 - val_loss: 0.7795 - val_accuracy: 0.7314
Epoch 27/30
274/274 [==============================] - 1s 5ms/step - loss: 0.6131 - accuracy: 0.7828 - val_loss: 0.7607 - val_accuracy: 0.7392
Epoch 28/30
274/274 [==============================] - 1s 5ms/step - loss: 0.5991 - accuracy: 0.7900 - val_loss: 0.7622 - val_accuracy: 0.7448
Epoch 29/30
274/274 [==============================] - 2s 6ms/step - loss: 0.5978 - accuracy: 0.7868 - val_loss: 0.7655 - val_accuracy: 0.7411
Epoch 30/30
274/274 [==============================] - 1s 5ms/step - loss: 0.5863 - accuracy: 0.7935 - val_loss: 0.7420 - val_accuracy: 0.7507

plt.plot(history.history['accuracy'], label='accuracy')
plt.plot(history.history['val_accuracy'], label = 'val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.ylim([0.3, 1])
plt.legend(loc='lower right')
plt.savefig("/kaggle/working/temp.png",dpi=800)

model.evaluate(test_images, test_labels)

313/313 [==============================] - 1s 2ms/step - loss: 0.7392 - accuracy: 0.7478
[0.7391936779022217, 0.7477999925613403]