Keras—猫狗数据集进行卷积（Conv2D）训练以及图像数据增强

 数据增强不可以增强验证集和测试集！！

# !/user/bin/env python
# -*- coding:utf-8 -*-
"""
@author:MXD
@file: 小型数据集训练卷积神经网络.py
@time: 2019/01/20 14:36
@software: PyCharm
"""
import os, shutil


###将数据分别存到各个文件夹
# 原始数据集的解压路径
originial_dataset_dir = 'D:\PythonPractical\KerasPra\download\kaggle_original_data'
base_dir = 'D:\PythonPractical\KerasPra\download\cats_and_dogs_small'
os.mkdir(base_dir)

train_dir = os.path.join(base_dir, 'train')
validation_dir = os.path.join(base_dir, 'validation')
os.mkdir(validation_dir)
test_dir = os.path.join(base_dir, 'test')
os.mkdir(test_dir)

train_cats_dir = os.path.join(base_dir, 'cats')
os.mkdir(train_cats_dir)
train_dogs_dir = os.path.join(base_dir, 'dogs')
os.mkdir(train_dogs_dir)

validation_cats_dir = os.path.join(validation_dir, 'cats')
os.mkdir(validation_cats_dir)
validation_dogs_dir = os.path.join(validation_dir, 'dogs')
os.mkdir(validation_dogs_dir)

test_cats_dir = os.path.join(test_dir, 'cats')
os.mkdir(test_cats_dir)
test_dogs_dir = os.path.join(test_dir, 'dogs')
os.mkdir(test_dogs_dir)

fnames = ["cat.{}.jpg".format(i) for i in range(1000)]
for fname in fnames:
    src = os.path.join(originial_dataset_dir, fname)
    dst = os.path.join(train_cats_dir, fname)
    shutil.copy(src, dst)
fnames = ["cat.{}.jpg".format(i) for i in range(1000, 1500)]
for fname in fnames:
    src = os.path.join(originial_dataset_dir, fname)
    dst = os.path.join(validation_cats_dir, fname)
    shutil.copy(src, dst)
fnames = ["cat.{}.jpg".format(i) for i in range(1500, 2000)]
for fname in fnames:
    src = os.path.join(originial_dataset_dir, fname)
    dst = os.path.join(test_cats_dir, fname)
    shutil.copy(src, dst)

fnames = ["dog.{}.jpg".format(i) for i in range(1000)]
for fname in fnames:
    src = os.path.join(originial_dataset_dir, fname)
    dst = os.path.join(train_dogs_dir, fname)
    shutil.copy(src, dst)
fnames = ["dog.{}.jpg".format(i) for i in range(1000, 1500)]
for fname in fnames:
    src = os.path.join(originial_dataset_dir, fname)
    dst = os.path.join(validation_dogs_dir, fname)
    shutil.copy(src, dst)
fnames = ["dog.{}.jpg".format(i) for i in range(1000, 1500)]
for fname in fnames:
    src = os.path.join(originial_dataset_dir, fname)
    dst = os.path.join(test_dogs_dir, fname)
    shutil.copy(src, dst)

### 构建网络
from keras import layers
from keras import models

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(150, 150, 3)))
model.add(layers.MaxPool2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPool2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPool2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPool2D((2, 2)))

model.add(layers.Flatten())
model.add(layers.Dropout(0.5)) #drop层：进一步降低过拟合
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()

from keras import optimizers

model.compile(loss='binarycrossentrop', optimizer=optimizers.RMSprop(lr=1e-4), metrics=['acc'])

# 使用ImageDataGenerator从目标中读取图像
from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale=1. / 255)  # 将所有图像乘以1/255缩放
test_datagen = ImageDataGenerator(rescale=1. / 255)

train_generator = train_datagen.flow_from_directory(
    train_dir, target_size=(150, 150), batch_size=20, class_mode='binary'
)  # 因为使用了binary_crossentropy损失，所以要使用二进制标签；target_size:指定图片的大小

validation_generator = test_datagen.flow_from_directory(
    validation_dir, target_size=(150, 150), batch_size=20, class_mode='binary')

history = model.fit_generator(train_generator, steps_per_epoch=100,
                              epochs=30,
                              validation_data=validation_generator,
                              validation_steps=50)
model.save('cats_and_dogs_small_1.h5')#训练完成保存模型是一种好习惯88

# 绘图
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, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()

plt.figure()
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
plt.show()

###数据增强
"""
数据增强是从现有的训练样本中生成更多的训练数据，其方法是利用多种能够生成可信图像的随机变换来增加（augment）样本。
其目标是，模型在训练时不会两次查看完全相同的图像。
这可以让模型能够观察导数据的更多内容，从而有更好的泛化能力。

注意：数据增强不能用于验证和测试数据集！！！

"""
from keras.preprocessing import image#图像预处理工具模块
from keras.preprocessing.image import ImageDataGenerator
# ImageDataGenerator：进行数据增强
datagen = ImageDataGenerator(
    rotation_range=40,#角度值（0——180范围内），表示图像随机旋转的角度范围
    width_shift_range=0.2,#width_shift,height_shift是图像在水平或垂直方向上平移的范围（相对于总宽度和总高度的比例）
    height_shift_range=0.2,
    shear_range=0.2,#是随机错切变换的角度
    zoom_range=0.2,#shear_range是图像缩放的范围
    horizontal_flip=True,#是随机将一半图像水平翻转。如果没有水平部队称的假设（比如真实世界的图像），这种做法是有意思的
    fill_mode='nearest'#用于填充新创建像素的方法，这些新像素可能来自于旋转或宽度/高度平移
)

fnames = [os.path.join(train_cats_dir,fname) for fname in os.listdir(train_cats_dir)]
img_path =fnames[3] #选择一个图片进行增强
img = image.load_img(img_path,target_size=(150.150))#读取图像并调整大小
x = image.img_to_array(img)#将其转化为形状（150,150,3）的numpy数组
x = x.reshape((1,)+x.shape)#将其形状改变为（1,150,150,3）
# 、生成随机变换后的图像批量。循环是无限的，因此你需要在某个时刻终止循环
i= 0
for batch in datagen.flow(x,batch_size=1):
    plt.figure(i)
    imgplot = plt.imshow(image.array_to_img(batch[0]))
    i+=1
    if i %4 ==0:
        break
plt.show()