class_3.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Keras入门课3：使用CNN识别cifar10数据集"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "cifar10是一个日常物品的数据集，一共有10类，属于是比较小的数据集。这次用一个4个卷积层加2个全连接层的典型CNN网络来进行分类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.datasets import cifar10\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "↓首先载入cifar10数据集，和mnist数据集的载入方法一致，本地没有数据的话会先下载"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "(x_train,y_train),(x_test,y_test) = cifar10.load_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "cifar10数据集图像大小是32*32的3通道彩图，训练集5万张，测试集1万张。和之前的mnist数据集不同，由于是彩色的，所以样本直接就是4维的。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(x_train.shape,y_train.shape)\n",
    "print(x_test.shape,y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.imshow(x_train[0])\n",
    "plt.show()\n",
    "plt.imshow(x_train[1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "可以看到数据读入没有问题，第一张是蛤蟆，第二张是一个卡车。\n",
    "\n",
    "↓规范化数据"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_train = x_train/255\n",
    "x_test = x_test/255\n",
    "y_train = keras.utils.to_categorical(y_train,10)\n",
    "y_test = keras.utils.to_categorical(y_test,10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "↓构建模型。之前构建模型都是先生成一个model，然后使用add方法来一层一层的加，现在用另一种更方便的方法。直接在Sequential初始化的时候按数组一个一个写进去就可以了。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential([\n",
    "    Conv2D(32,(3,3),padding='same',input_shape=(32,32,3),activation='relu'),\n",
    "    Conv2D(32,(3,3),activation='relu'),\n",
    "    MaxPooling2D(pool_size=(2,2)),\n",
    "    Dropout(0.25),\n",
    "    \n",
    "    Conv2D(64,(3,3),padding='same',activation='relu'),\n",
    "    Conv2D(64,(3,3),activation='relu'),\n",
    "    MaxPooling2D(pool_size=(2,2)),\n",
    "    Dropout(0.25),\n",
    "    \n",
    "    Flatten(),\n",
    "    Dense(512,activation='relu'),\n",
    "    Dropout(0.5),\n",
    "    Dense(10,activation='softmax')    \n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "↓指定优化函数的参数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "opt = keras.optimizers.rmsprop(lr=0.0001,decay=1e-6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "             optimizer=opt,\n",
    "             metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 至此直接调用fit方法就可以进行训练了。但是为了模型更快的收敛以及更好的泛化性能，往往我们会对图像做一些变换，比如缩放、平移、旋转等等。下面我们要用keras自带的图像增强来对图像做一些变换"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "↓这里生成了一个数据增强器，包含了范围20°内的随机旋转，±15%的缩放以及随机的水平翻转。可调的参数还有很多，具体的可以查看文档。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "datagen = ImageDataGenerator(\n",
    "    rotation_range = 20,\n",
    "    zoom_range = 0.15,\n",
    "    horizontal_flip = True,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# datagen.fit(x_train) 只有使用featurewise_center，featurewise_std_normalization或zca_whitening时需要此函数"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "↓通过ImageDataGenerator生成的数据需要使用model的fit_generator方法来进行训练，其中的workers参数表示多线程运算。\n",
    "\n",
    "datagen的flow方法可以按批次的生成训练所需数据，注意这里生成的数据都是经过了数据增强的，并且是实时的。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.fit_generator(datagen.flow(x_train,y_train,batch_size=64),steps_per_epoch = 1000,epochs = 2,\n",
    "                    validation_data=(x_test,y_test),workers=4,verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "↓保存模型，包括了模型的结构以及参数。后缀用h5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('cifar10_trained_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "scores = model.evaluate(x_test,y_test,verbose=1)\n",
    "print('Test loss:',scores[0])\n",
    "print('Test accuracy:',scores[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 总结\n",
    "\n",
    "1. 学习了一种新的使用Sequential()构建模型的方法，更加的简单快捷\n",
    "1. 学习了使用Keras内置的ImageDataGenerator来做数据增强的方法\n",
    "1. 调用model的fit_generator来进行针对增强数据的训练\n",
    "1. 学习了如何保存模型\n",
    "\n",
    "本文代码链接：https://github.com/tsycnh/Keras-Tutorials/blob/master/class_3.ipynb\n",
    "\n",
    "参考\n",
    "> https://github.com/keras-team/keras/blob/master/examples\n",
    "> https://keras-cn.readthedocs.io/en/latest/preprocessing/image/"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}