第五章 深度学习用于计算机视觉

""" 3.1.1.使用预训练的卷积神经网络(不使用数据增强的快速特征提取) """
# 1.将VGG16卷积基实例化
from keras.applications import VGG16        # conv_base是一个VGG模型

conv_base = VGG16(weights = 'imagenet',          # 指定模型初始化的权重检查点
                  include_top = False,           # 是否包含密集连接分类器
                  input_shape = (150,150, 3))


# 2.使用预训练的卷积基提取特征(2-1：不使用数据增强的快速特征提取)
import os
import numpy as np
from keras.preprocessing.image import ImageDataGenerator

base_dir = 'F:\datasets\dogs-vs-csts\猫狗大战'            # 保存较小数据集的目录
train_dir = os.path.join(base_dir, 'train')              # 划分后的训练集
validation_dir = os.path.join(base_dir, 'validation')    # 划分后的验证集
test_dir = os.path.join(base_dir, 'test')                # 划分后的测试集

datagen = ImageDataGenerator(rescale=1./255)     # 将所有图像乘以1/255缩放
batch_size = 20                                  # 小批量


def extract_features(directory, sample_count):                  # 提取_特征(目录，样本数)
    features = np.zeros(shape=(sample_count, 4, 4, 512))        # 4D (样本数, 4，4，512)
    labels = np.zeros(shape=(sample_count))                     # 2000个0
    
    # 以文件夹路径为参数,生成经过数据提升/归一化后的数据,在一个无限循环中无限产生batch数据
    generator = datagen.flow_from_directory(       # 读取图像的生成器
        directory,                  # 目录
        target_size = (150,150),    # 整数tuple元组,默认为(256, 256). 图像将被resize成该尺寸
        batch_size = batch_size,    # batch数据的大小，默认32
        class_mode = 'binary')      # 颜色模式
    
    i = 0
    # 生成器：下面将每个批次的输出结果通过conv_pridect()写到特征矩阵中等待接下来的输入全连接层处理。
    for inputs_batch, labels_batch in generator:             
        features_batch = conv_base.predict(inputs_batch)      # 获得输出结果用的是predict方法
        # 提取特征数据是在将数据先输入conv_base后得到的结果数据，然后再将其保存为numpy数组
        features[i * batch_size : (i+1) * batch_size] = features_batch
        labels[i * batch_size : (i+1) * batch_size] = labels_batch
        i += 1
        if i*batch_size >= sample_count:
            break       # 注意,这些生成器在循环中不断生成数据，所以必须在读取完所有图像后终止循环
    return features, labels


train_features, train_labels = extract_features(train_dir, 2000)        # 提取的特征形状为:4D   
validation_features, validation_labels = extract_features(validation_dir, 1000)
test_features, test_labels = extract_features(test_dir, 1000)

train_features = np.reshape(train_features, (2000, 4*4*512))           # 将特征其展平为：2D   将数据铺平
validation_features = np.reshape(validation_features, (1000, 4*4*512))
test_features = np.reshape(test_features, (1000, 4*4*512))


# 3.定义并训练密集连接分类器
from keras import models
from keras import layers
from keras import optimizers

model = models.Sequential()
model.add(layers.Dense(256, activation='relu', input_dim=4*4*512))
model.add(layers.Dropout(0.5))         # Dropout正则化
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()

model.compile(optimizer = optimizers.RMSprop(lr=2e-5),    # 配置模型
              loss = 'binary_crossentropy',
              metrics = ['acc'])

history = model.fit(train_features, train_labels,         # 训练数据
                    epochs = 30,
                    batch_size = 20,
                    validation_data = (validation_features, validation_labels))     # 验证数据

loss,accuracy = model.evaluate(test_features, test_labels)        # 评估测试集

# 4.绘制训练过程中的损失曲线和精度曲线
import matplotlib.pyplot as plt

acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1, len(acc)+1)

plt.plot(epochs, acc, 'r', label='Training acc')            # 绘制训练过程中的精度曲线
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and Validation accuracy')
plt.legend()

plt.figure()

plt.plot(epochs, loss, 'r', label='Training loss')          # 绘制训练过程中的损失曲线
plt.plot(epochs, val_loss, 'b', label='Validation val_loss')
plt.title('Training and Validation loss')
plt.legend()
plt.show()