365天深度学习训练营-第2周：彩色图像分类

第二周：彩色图像分类

• 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
• 🍦 参考文章地址： 🔗深度学习100例-卷积神经网络（CNN）彩色图片分类 | 第2天
• 🍖 作者：K同学啊

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.cifar10.load_data()

归一化

# 将像素的值标准化到0-1之间
train_images, test_images = train_images / 255.0, test_images / 255.0
print('train_images.shape:', train_images.shape)
print('test_images.shape:', test_images.shape)
print('train_labels.shape:', train_labels.shape)
print('test_labels.shape:', test_labels.shape)

结果

可视化展示

class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer','dog', 'frog', 'horse', 'ship', 'truck']

plt.figure(figsize=(20,10))
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(class_names[train_labels[i][0]])
    plt.tight_layout()
plt.show()

构建CNN网络

model = models.Sequential([
    layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)),  # 卷积层1，卷积核3*3
    layers.MaxPooling2D((2, 2)),  # 池化层1，2*2采样
    layers.Conv2D(64, (3, 3), activation='relu'),  # 卷积层2，卷积核3*3
    layers.MaxPooling2D((2, 2)),  # 池化层2，2*2采样
    layers.Conv2D(64, (3, 3), activation='relu'),  # 卷积层3，卷积核3*3
    layers.Dropout(.2),


    layers.Flatten(),  # Flatten层，连接卷积层与全连接层
    layers.Dense(128, activation='relu'),  # 全连接层，特征进一步提取
    layers.Dropout(.2),
    layers.Dense(64, activation='relu'),  # 全连接层，特征进一步提取
    layers.Dense(10)  # 输出层，输出预期结果"
])

model.summary()  # 打印网络结构

模型的编译

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

模型训练

history = model.fit(train_images, train_labels, epochs=10, 
                    validation_data=(test_images, test_labels))

Epoch 1/10
1563/1563 [==============================] - 66s 40ms/step - loss: 1.7866 - accuracy: 0.3319 - val_loss: 1.2316 - val_accuracy: 0.5562
Epoch 2/10
1563/1563 [==============================] - 60s 39ms/step - loss: 1.2366 - accuracy: 0.5546 - val_loss: 1.0473 - val_accuracy: 0.6315
Epoch 3/10
1563/1563 [==============================] - 59s 38ms/step - loss: 1.0551 - accuracy: 0.6243 - val_loss: 0.9543 - val_accuracy: 0.6647
Epoch 4/10
1563/1563 [==============================] - 52s 34ms/step - loss: 0.9662 - accuracy: 0.6582 - val_loss: 0.9473 - val_accuracy: 0.6646
Epoch 5/10
1563/1563 [==============================] - 53s 34ms/step - loss: 0.8798 - accuracy: 0.6882 - val_loss: 0.8572 - val_accuracy: 0.7010
Epoch 6/10
1563/1563 [==============================] - 53s 34ms/step - loss: 0.8260 - accuracy: 0.7094 - val_loss: 0.8749 - val_accuracy: 0.6918
Epoch 7/10
1563/1563 [==============================] - 53s 34ms/step - loss: 0.7827 - accuracy: 0.7210 - val_loss: 0.8275 - val_accuracy: 0.7105
Epoch 8/10
1563/1563 [==============================] - 54s 35ms/step - loss: 0.7533 - accuracy: 0.7318 - val_loss: 0.7889 - val_accuracy: 0.7218
Epoch 9/10
1563/1563 [==============================] - 53s 34ms/step - loss: 0.7107 - accuracy: 0.7516 - val_loss: 0.8129 - val_accuracy: 0.7186
Epoch 10/10
1563/1563 [==============================] - 53s 34ms/step - loss: 0.6797 - accuracy: 0.7609 - val_loss: 0.7837 - val_accuracy: 0.7309

预测

plt.imshow(test_images[1])
plt.show()

pre = model.predict(test_images)
print(class_names[np.argmax(pre[1])])

ship

模型评估

plt.plot(history.history['accuracy'], label='accuracy')
plt.plot(history.history['val_accuracy'], label='val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.ylim([0.5, 1])
plt.legend(loc='lower right')
plt.show()
test_loss, test_acc = model.evaluate(test_images,  test_labels, verbose=2)

313/313 - 3s - loss: 0.7837 - accuracy: 0.7309

print(test_acc)

0.73089998960495

​