人工智能学习：CIFAR-10数据分类识别-VGG网络（5）

星光2020

已于 2022-11-14 11:48:22 修改

阅读量1.2k

点赞数

分类专栏：人工智能文章标签：人工智能 CIFAR-10 VGG

于 2022-11-07 02:57:23 首次发布

本文链接：https://blog.csdn.net/mbdong/article/details/127724152

版权

人工智能专栏收录该内容

25 篇文章 5 订阅

订阅专栏

这里尝试采用VGG网络对CIFAR-10数据集进行分类识别。

1 导入需要的模块

import numpy as np

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

import matplotlib.pyplot as plt

2 载入CIFAR-10数据集

# load CIFAR-10 dataset
(train_images, train_labels), (test_images, test_labels) = keras.datasets.cifar10.load_data()
# train_images: 50000*32*32*3, train_labels: 50000*1, test_images: 10000*32*32*3, test_labels: 10000*1

# change data shape & types
train_input = train_images/255.0
test_input = test_images/255.0
train_output = train_labels
test_output = test_labels

3 构建神经网络

首先，定义构建模型函数

def build_model():
    model = models.Sequential()
    
    # 1st layer, input shape (32,32,3)
    model.add(layers.Conv2D(64, (3,3), padding='same', input_shape=(32,32,3)))
    model.add(layers.Activation('relu'))
    model.add(layers.BatchNormalization())
    model.add(layers.Dropout(0.3))
    
    # 2nd layer, input shape (32,32,64)
    model.add(layers.Conv2D(64, (3,3), padding='same'))
    model.add(layers.Activation('relu'))
    model.add(layers.BatchNormalization())
    model.add(layers.MaxPooling2D(pool_size=(2,2), strides=(2,2), padding='same'))

    # 3rd layer, (16,16,64)
    model.add(layers.Conv2D(128, (3,3), padding='same'))
    model.add(layers.Activation('relu'))
    model.add(layers.BatchNormalization())
    model.add(layers.Dropout(0.4))
              
    # 4th layer, (16,16,128)
    model.add(layers.Conv2D(128, (3,3), padding='same'))
    model.add(layers.Activation('relu'))
    model.add(layers.BatchNormalization())
    model.add(layers.MaxPooling2D(pool_size=(2,2)))
    
    # 5th layer, (8,8,128)
    model.add(layers.Conv2D(256, (3, 3), padding='same'))
    model.add(layers.Activation('relu'))
    model.add(layers.BatchNormalization())
    model.add(layers.Dropout(0.4))

    # 6th layer, (8,8,256)
    model.add(layers.Flatten())
    model.add(layers.Dense(512))
    model.add(layers.Activation('relu'))
    model.add(layers.BatchNormalization())
    
    #7th layer, 512
    model.add(layers.Dropout(0.5))
    model.add(layers.Dense(10))
    model.add(layers.Activation('softmax'))
    
        
    model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['sparse_categorical_accuracy'])
    
    return model

这里构建5层卷积层，加上2层全连接层。调用函数

# build model
network = build_model()

# show network summary
network.summary()

显示结果如下

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 32, 32, 64)        1792      
_________________________________________________________________
activation (Activation)      (None, 32, 32, 64)        0         
_________________________________________________________________
batch_normalization (BatchNo (None, 32, 32, 64)        256       
_________________________________________________________________
dropout (Dropout)            (None, 32, 32, 64)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 32, 32, 64)        36928     
_________________________________________________________________
activation_1 (Activation)    (None, 32, 32, 64)        0         
_________________________________________________________________
batch_normalization_1 (Batch (None, 32, 32, 64)        256       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 16, 16, 64)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 16, 16, 128)       73856     
_________________________________________________________________
activation_2 (Activation)    (None, 16, 16, 128)       0         
_________________________________________________________________
batch_normalization_2 (Batch (None, 16, 16, 128)       512       
_________________________________________________________________
dropout_1 (Dropout)          (None, 16, 16, 128)       0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 16, 16, 128)       147584    
_________________________________________________________________
activation_3 (Activation)    (None, 16, 16, 128)       0         
_________________________________________________________________
batch_normalization_3 (Batch (None, 16, 16, 128)       512       
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 8, 8, 128)         0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 8, 8, 256)         295168    
_________________________________________________________________
activation_4 (Activation)    (None, 8, 8, 256)         0         
_________________________________________________________________
batch_normalization_4 (Batch (None, 8, 8, 256)         1024      
_________________________________________________________________
dropout_2 (Dropout)          (None, 8, 8, 256)         0         
_________________________________________________________________
flatten (Flatten)            (None, 16384)             0         
_________________________________________________________________
dense (Dense)                (None, 512)               8389120   
_________________________________________________________________
activation_5 (Activation)    (None, 512)               0         
_________________________________________________________________
batch_normalization_5 (Batch (None, 512)               2048      
_________________________________________________________________
dropout_3 (Dropout)          (None, 512)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 10)                5130      
_________________________________________________________________
activation_6 (Activation)    (None, 10)                0         
=================================================================
Total params: 8,954,186
Trainable params: 8,951,882
Non-trainable params: 2,304
_________________________________________________________________

4 训练模型

调用函数训练

# train model
history = network.fit(train_input, train_output, epochs=30, batch_size=256, validation_split=0.1)

训练30次，batch_size为256，训练结果显示如下

Epoch 1/30
176/176 [==============================] - 15s 62ms/step - loss: 1.7285 - sparse_categorical_accuracy: 0.4502 - val_loss: 4.3244 - val_sparse_categorical_accuracy: 0.1566
Epoch 2/30
176/176 [==============================] - 9s 53ms/step - loss: 1.0967 - sparse_categorical_accuracy: 0.6184 - val_loss: 4.2188 - val_sparse_categorical_accuracy: 0.2258
Epoch 3/30
176/176 [==============================] - 9s 53ms/step - loss: 0.8392 - sparse_categorical_accuracy: 0.7048 - val_loss: 1.5962 - val_sparse_categorical_accuracy: 0.5262
Epoch 4/30
176/176 [==============================] - 9s 53ms/step - loss: 0.6949 - sparse_categorical_accuracy: 0.7549 - val_loss: 0.7939 - val_sparse_categorical_accuracy: 0.7430
Epoch 5/30
176/176 [==============================] - 9s 53ms/step - loss: 0.5994 - sparse_categorical_accuracy: 0.7880 - val_loss: 0.9743 - val_sparse_categorical_accuracy: 0.7168
Epoch 6/30
176/176 [==============================] - 9s 54ms/step - loss: 0.5187 - sparse_categorical_accuracy: 0.8173 - val_loss: 0.8175 - val_sparse_categorical_accuracy: 0.7520
Epoch 7/30
176/176 [==============================] - 9s 54ms/step - loss: 0.4544 - sparse_categorical_accuracy: 0.8398 - val_loss: 0.8302 - val_sparse_categorical_accuracy: 0.7544
Epoch 8/30
176/176 [==============================] - 9s 54ms/step - loss: 0.3901 - sparse_categorical_accuracy: 0.8629 - val_loss: 0.7184 - val_sparse_categorical_accuracy: 0.7934
Epoch 9/30
176/176 [==============================] - 9s 54ms/step - loss: 0.3390 - sparse_categorical_accuracy: 0.8794 - val_loss: 0.8141 - val_sparse_categorical_accuracy: 0.7962
Epoch 10/30
176/176 [==============================] - 10s 54ms/step - loss: 0.2886 - sparse_categorical_accuracy: 0.8964 - val_loss: 0.9829 - val_sparse_categorical_accuracy: 0.7804
Epoch 11/30
176/176 [==============================] - 10s 54ms/step - loss: 0.2630 - sparse_categorical_accuracy: 0.9075 - val_loss: 0.7088 - val_sparse_categorical_accuracy: 0.8034
Epoch 12/30
176/176 [==============================] - 10s 54ms/step - loss: 0.2362 - sparse_categorical_accuracy: 0.9164 - val_loss: 0.5813 - val_sparse_categorical_accuracy: 0.8336
Epoch 13/30
176/176 [==============================] - 10s 54ms/step - loss: 0.2086 - sparse_categorical_accuracy: 0.9269 - val_loss: 0.7702 - val_sparse_categorical_accuracy: 0.8014
Epoch 14/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1860 - sparse_categorical_accuracy: 0.9345 - val_loss: 0.7444 - val_sparse_categorical_accuracy: 0.8254
Epoch 15/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1748 - sparse_categorical_accuracy: 0.9398 - val_loss: 0.7130 - val_sparse_categorical_accuracy: 0.8184
Epoch 16/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1582 - sparse_categorical_accuracy: 0.9443 - val_loss: 0.7712 - val_sparse_categorical_accuracy: 0.8226
Epoch 17/30
176/176 [==============================] - 10s 55ms/step - loss: 0.1459 - sparse_categorical_accuracy: 0.9488 - val_loss: 0.8808 - val_sparse_categorical_accuracy: 0.8086
Epoch 18/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1329 - sparse_categorical_accuracy: 0.9530 - val_loss: 0.7062 - val_sparse_categorical_accuracy: 0.8340
Epoch 19/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1323 - sparse_categorical_accuracy: 0.9538 - val_loss: 0.6216 - val_sparse_categorical_accuracy: 0.8380
Epoch 20/30
176/176 [==============================] - 10s 55ms/step - loss: 0.1243 - sparse_categorical_accuracy: 0.9575 - val_loss: 0.6749 - val_sparse_categorical_accuracy: 0.8334
Epoch 21/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1206 - sparse_categorical_accuracy: 0.9586 - val_loss: 0.7408 - val_sparse_categorical_accuracy: 0.8268
Epoch 22/30
176/176 [==============================] - 10s 54ms/step - loss: 0.1105 - sparse_categorical_accuracy: 0.9615 - val_loss: 0.7999 - val_sparse_categorical_accuracy: 0.8314
Epoch 23/30
176/176 [==============================] - 10s 55ms/step - loss: 0.1064 - sparse_categorical_accuracy: 0.9633 - val_loss: 0.6867 - val_sparse_categorical_accuracy: 0.8396
Epoch 24/30
176/176 [==============================] - 10s 54ms/step - loss: 0.0974 - sparse_categorical_accuracy: 0.9655 - val_loss: 0.6695 - val_sparse_categorical_accuracy: 0.8422
Epoch 25/30
176/176 [==============================] - 10s 54ms/step - loss: 0.0908 - sparse_categorical_accuracy: 0.9687 - val_loss: 0.7222 - val_sparse_categorical_accuracy: 0.8306
Epoch 26/30
176/176 [==============================] - 10s 54ms/step - loss: 0.0907 - sparse_categorical_accuracy: 0.9689 - val_loss: 0.6841 - val_sparse_categorical_accuracy: 0.8384
Epoch 27/30
176/176 [==============================] - 10s 55ms/step - loss: 0.0866 - sparse_categorical_accuracy: 0.9696 - val_loss: 0.8356 - val_sparse_categorical_accuracy: 0.8286
Epoch 28/30
176/176 [==============================] - 10s 55ms/step - loss: 0.0898 - sparse_categorical_accuracy: 0.9690 - val_loss: 0.6899 - val_sparse_categorical_accuracy: 0.8392
Epoch 29/30
176/176 [==============================] - 10s 55ms/step - loss: 0.0867 - sparse_categorical_accuracy: 0.9700 - val_loss: 0.7572 - val_sparse_categorical_accuracy: 0.8338
Epoch 30/30
176/176 [==============================] - 10s 55ms/step - loss: 0.0796 - sparse_categorical_accuracy: 0.9728 - val_loss: 0.7699 - val_sparse_categorical_accuracy: 0.8336

经过训练，得到0.9728的训练精度和0.8336的测试精度。对训练过程进行绘制，如下

# plot train history
loss = history.history['loss']
val_loss = history.history['val_loss']
acc = history.history['sparse_categorical_accuracy']
val_acc = history.history['val_sparse_categorical_accuracy']

plt.figure(figsize=(10,3))

plt.subplot(1,2,1)
plt.plot(loss, color='blue', label='train')
plt.plot(val_loss, color='red', label='test')
plt.ylabel('loss')
plt.legend()

plt.subplot(1,2,2)
plt.plot(acc, color='blue', label='train')
plt.plot(val_acc, color='red', label='test')
plt.ylabel('accuracy')
plt.legend()

显示如下的结果
在这里插入图片描述
训练和测试的准确度和损失函数，都随着训练次数的增加，逐渐优化。显示在泛化能力上要好于之前的模型。

5 测试训练模型

# evaluate model
network.evaluate(test_input, test_output, verbose=2)

显示在测试集上的准确度和损失函数

313/313 - 1s - loss: 0.7931 - sparse_categorical_accuracy: 0.8315
[0.7930535078048706, 0.8314999938011169]

结果和训练给出的性能指标接近。绘制测试集前100张图片的测试结果

# predict on test data
predict_output = network.predict(test_input)

# lines and columns of subplots
m = 10
n = 10
num = m*n

# labels of category
labels = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']

# figure size
plt.figure(figsize=(15,15))

# plot first 100 pictures and results in test images
for i in range(num):
    plt.subplot(m,n,i+1)
    
    type_index = np.argmax(predict_output[i]);
    label = labels[type_index]
    
    clr = 'black' if type_index == test_labels[i] else 'red'
                 
    plt.imshow(test_images[i])
    #plt.axis('off')
    plt.xticks([])
    plt.yticks([])
    
    plt.xlabel(label, color=clr)

plt.show()