4-Write-your-own-dataset-class_ja.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# データセットクラスを書いてみよう\n",
    "\n",
    "ここでは、Chainerにすでに用意されているCIFAR10のデータを取得する機能を使って、データセットクラスを自分で書いてみます。Chainerでは、データセットを表すクラスは以下の機能を持っていることが必要とされます。\n",
    "\n",
    "- データセット内のデータ数を返す`__len__`メソッド\n",
    "- 引数として渡される`i`に対応したデータもしくはデータとラベルの組を返す`get_example`メソッド\n",
    "\n",
    "その他のデータセットに必要な機能は、`chainer.dataset.DatasetMixin`クラスを継承することで用意できます。ここでは、`DatasetMixin`クラスを継承し、Data augmentation機能のついたデータセットクラスを作成してみましょう。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## 1. CIFAR10データセットクラスを書く"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from chainer import dataset\n",
    "from chainer.datasets import cifar\n",
    "\n",
    "class CIFAR10Augmented(dataset.DatasetMixin):\n",
    "\n",
    "    def __init__(self, train=True):\n",
    "        train_data, test_data = cifar.get_cifar10()\n",
    "        if train:\n",
    "            self.data = train_data\n",
    "        else:\n",
    "            self.data = test_data\n",
    "        self.train = train\n",
    "        self.random_crop = 4\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.data)\n",
    "\n",
    "    def get_example(self, i):\n",
    "        x, t = self.data[i]\n",
    "        if self.train:\n",
    "            x = x.transpose(1, 2, 0)\n",
    "            h, w, _ = x.shape\n",
    "            x_offset = np.random.randint(self.random_crop)\n",
    "            y_offset = np.random.randint(self.random_crop)\n",
    "            x = x[y_offset:y_offset + h - self.random_crop,\n",
    "                  x_offset:x_offset + w - self.random_crop]\n",
    "            if np.random.rand() > 0.5:\n",
    "                x = np.fliplr(x)\n",
    "            x = x.transpose(2, 0, 1)\n",
    "        return x, t"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "このクラスは、CIFAR10のデータのそれぞれに対し、\n",
    "\n",
    "- 32x32の大きさの中からランダムに28x28の領域をクロップ\n",
    "- 1/2の確率で左右を反転させる\n",
    "\n",
    "という加工を行っています。こういった操作を加えることで擬似的に学習データのバリエーションを増やすと、オーバーフィッティングを抑制することに役に立つということが知られています。これらの操作以外にも、画像の色味を変化させるような変換やランダムな回転、アフィン変換など、さまざまな加工によって学習データ数を擬似的に増やす方法が提案されています。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## 2. 作成したデータセットクラスを使って学習を行う\n",
    "\n",
    "それではさっそくこの`CIFAR10`クラスを使って学習を行ってみましょう。先程使ったのと同じ大きなネットワークを使うことで、Data augmentationの効果がどの程度あるのかを調べてみましょう。`train`関数も含め、データセットクラス以外は先程使用したコードとほぼ同じになっています。異なるところはエポック数と、保存先ディレクトリ名だけです。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch       main/loss   main/accuracy  validation/main/loss  validation/main/accuracy  elapsed_time\n",
      "\u001b[J1           2.023       0.248981       1.75221               0.322353                  18.4387       \n",
      "\u001b[J2           1.51639     0.43716        1.36708               0.512639                  36.482        \n",
      "\u001b[J3           1.25354     0.554177       1.17713               0.586087                  54.6892       \n",
      "\u001b[J4           1.05922     0.637804       0.971438              0.665904                  72.9602       \n",
      "\u001b[J5           0.895339    0.701886       0.918005              0.706409                  91.4061       \n",
      "\u001b[J6           0.783455    0.741717       0.766948              0.745422                  109.907       \n",
      "\u001b[J7           0.700934    0.771947       0.720493              0.759952                  128.419       \n",
      "\u001b[J8           0.634291    0.792013       0.643777              0.789112                  146.979       \n",
      "\u001b[J9           0.582722    0.808724       0.599772              0.807424                  165.56        \n",
      "\u001b[J10          0.546745    0.821863       0.567675              0.818969                  184.099       \n",
      "\u001b[J11          0.504822    0.833107       0.544959              0.825836                  202.647       \n",
      "\u001b[J12          0.485791    0.841589       0.503365              0.836684                  221.199       \n",
      "\u001b[J13          0.457064    0.849884       0.51921               0.830713                  239.778       \n",
      "\u001b[J14          0.438434    0.856734       0.499945              0.838177                  258.384       \n",
      "\u001b[J15          0.407408    0.866997       0.523641              0.837679                  276.959       \n",
      "\u001b[J16          0.388121    0.872519       0.474266              0.847631                  295.473       \n",
      "\u001b[J17          0.38036     0.87516        0.498298              0.851413                  314.007       \n",
      "\u001b[J18          0.361475    0.883403       0.478719              0.855494                  332.557       \n",
      "\u001b[J19          0.342159    0.888244       0.478004              0.855394                  351.049       \n",
      "\u001b[J20          0.331293    0.891565       0.45817               0.857086                  369.542       \n",
      "\u001b[J21          0.323042    0.894781       0.464906              0.862361                  388.108       \n",
      "\u001b[J22          0.308934    0.898227       0.441302              0.865446                  406.624       \n",
      "\u001b[J23          0.291801    0.903889       0.482271              0.869128                  425.1         \n",
      "\u001b[J24          0.287181    0.90511        0.513744              0.862261                  443.59        \n",
      "\u001b[J25          0.283392    0.907509       0.445071              0.869725                  462.152       \n",
      "\u001b[J26          0.27007     0.911172       0.483729              0.860768                  480.68        \n",
      "\u001b[J27          0.262051    0.915013       0.468225              0.870521                  499.177       \n",
      "\u001b[J28          0.257783    0.916093       0.483101              0.875                     517.718       \n",
      "\u001b[J29          0.246585    0.920197       0.425248              0.876095                  536.226       \n",
      "\u001b[J30          0.234108    0.923115       0.472583              0.866441                  554.713       \n",
      "\u001b[J31          0.229523    0.924896       0.508774              0.874204                  573.227       \n",
      "\u001b[J32          0.222134    0.926937       0.516485              0.874104                  591.717       \n",
      "\u001b[J33          0.223208    0.92707        0.459016              0.87918                   610.234       \n",
      "\u001b[J34          0.209783    0.930758       0.483541              0.878583                  628.721       \n",
      "\u001b[J35          0.214651    0.930058       0.460262              0.872412                  647.251       \n",
      "\u001b[J36          0.200795    0.934079       0.483824              0.87719                   665.737       \n",
      "\u001b[J37          0.199789    0.934103       0.444633              0.882763                  684.246       \n",
      "\u001b[J38          0.192346    0.93736        0.568936              0.869526                  702.778       \n",
      "\u001b[J39          0.188353    0.9389         0.487995              0.881967                  721.257       \n",
      "\u001b[J40          0.18422     0.93908        0.526045              0.876194                  739.853       \n",
      "\u001b[J41          0.185171    0.939918       0.53164               0.876493                  758.578       \n",
      "\u001b[J42          0.17834     0.941421       0.489131              0.888137                  777.118       \n",
      "\u001b[J43          0.173166    0.943622       0.525443              0.883161                  795.586       \n",
      "\u001b[J44          0.162396    0.946303       0.501248              0.884057                  814.063       \n",
      "\u001b[J45          0.170292    0.944713       0.498577              0.882663                  832.589       \n",
      "\u001b[J46          0.16108     0.947823       0.509704              0.882763                  851.071       \n",
      "\u001b[J47          0.157382    0.948644       0.549481              0.888236                  869.543       \n",
      "\u001b[J48          0.157711    0.948944       0.474189              0.886545                  888.066       \n",
      "\u001b[J49          0.149987    0.950727       0.501718              0.884156                  906.556       \n",
      "\u001b[J50          0.15049     0.952045       0.571395              0.88744                   925.035       \n",
      "\u001b[J51          0.144148    0.952985       0.530693              0.881369                  943.545       \n",
      "\u001b[J52          0.142236    0.953325       0.499273              0.890227                  962.016       \n",
      "\u001b[J53          0.139191    0.954683       0.536683              0.89381                   980.517       \n",
      "\u001b[J54          0.141176    0.954205       0.601934              0.890326                  998.983       \n",
      "\u001b[J55          0.136463    0.955726       0.524247              0.88754                   1017.5        \n",
      "\u001b[J56          0.133722    0.956126       0.540561              0.89162                   1035.98       \n",
      "\u001b[J57          0.12951     0.95858        0.472836              0.889331                  1054.46       \n",
      "\u001b[J58          0.128031    0.958387       0.514021              0.889132                  1072.97       \n",
      "\u001b[J59          0.128484    0.959067       0.560094              0.890625                  1091.55       \n",
      "\u001b[J60          0.124952    0.959527       0.643559              0.886545                  1110.06       \n",
      "\u001b[J61          0.126153    0.959839       0.547355              0.887639                  1128.61       \n",
      "\u001b[J62          0.131744    0.959787       0.581561              0.891819                  1147.09       \n",
      "\u001b[J63          0.119843    0.961188       0.592228              0.887838                  1165.57       \n",
      "\u001b[J64          0.113508    0.963808       0.572284              0.887639                  1184.04       \n",
      "\u001b[J65          0.126285    0.959639       0.600794              0.880374                  1202.59       \n",
      "\u001b[J66          0.111629    0.964689       0.593115              0.892317                  1221.08       \n",
      "\u001b[J67          0.113058    0.964869       0.720929              0.89381                   1239.56       \n",
      "\u001b[J68          0.108513    0.964829       0.701805              0.884256                  1258.08       \n",
      "\u001b[J69          0.109513    0.965114       0.565101              0.889729                  1276.59       \n",
      "\u001b[J70          0.110135    0.964989       0.545945              0.888635                  1295.08       \n",
      "\u001b[J71          0.108055    0.966009       0.678656              0.890525                  1313.59       \n",
      "\u001b[J72          0.10718     0.964929       0.709935              0.890625                  1332.07       \n",
      "\u001b[J73          0.104518    0.966292       0.619641              0.889232                  1350.67       \n",
      "\u001b[J74          0.106752    0.965609       0.617278              0.889132                  1369.28       \n",
      "\u001b[J75          0.101721    0.96717        0.62095               0.891023                  1387.94       \n",
      "\u001b[J76          0.0992725   0.96863        0.613595              0.891521                  1406.44       \n",
      "\u001b[J77          0.101973    0.967451       0.660645              0.893113                  1424.95       \n",
      "\u001b[J78          0.0994073   0.96865        0.607272              0.895203                  1443.49       \n",
      "\u001b[J79          0.0978486   0.96901        0.879328              0.88953                   1461.98       \n",
      "\u001b[J80          0.0949811   0.96975        0.748834              0.891023                  1480.48       \n",
      "\u001b[J81          0.0966707   0.970189       0.705469              0.890924                  1499.03       \n",
      "\u001b[J82          0.0919403   0.970351       0.610847              0.889232                  1517.52       \n",
      "\u001b[J83          0.0950202   0.97017        0.747532              0.89172                   1536.02       \n",
      "\u001b[J84          0.0906361   0.970611       0.554294              0.88953                   1554.49       \n",
      "\u001b[J85          0.094787    0.970249       0.57539               0.890127                  1573.04       \n",
      "\u001b[J86          0.0918775   0.971011       0.790777              0.887241                  1591.5        \n",
      "\u001b[J87          0.0965973   0.96957        0.691885              0.894307                  1609.96       \n",
      "\u001b[J88          0.0921529   0.970851       0.71511               0.890028                  1628.48       \n",
      "\u001b[J89          0.0872026   0.972546       0.723237              0.892715                  1646.97       \n",
      "\u001b[J90          0.0875947   0.971831       0.736676              0.894606                  1665.45       \n",
      "\u001b[J91          0.0904918   0.971331       0.717111              0.883857                  1683.94       \n",
      "\u001b[J92          0.0864843   0.972071       0.627684              0.893611                  1702.4        \n",
      "\u001b[J93          0.0871864   0.973006       0.555012              0.892815                  1720.89       \n",
      "\u001b[J94          0.0869806   0.973311       0.730194              0.890824                  1739.36       \n",
      "\u001b[J95          0.0877855   0.973171       0.726305              0.89162                   1757.87       \n",
      "\u001b[J96          0.0780378   0.976012       0.943201              0.890725                  1776.41       \n",
      "\u001b[J97          0.086231    0.973765       0.57783               0.890227                  1794.99       \n",
      "\u001b[J98          0.0869593   0.973512       1.65576               0.878981                  1813.52       \n",
      "\u001b[J99          0.0870466   0.972931       0.718033              0.891421                  1831.99       \n",
      "\u001b[J100         0.079011    0.975332       0.754114              0.892815                  1850.46       \n"
     ]
    }
   ],
   "source": [
    "import chainer\n",
    "import chainer.functions as F\n",
    "import chainer.links as L\n",
    "from chainer.datasets import cifar\n",
    "from chainer import iterators\n",
    "from chainer import optimizers\n",
    "from chainer import training\n",
    "from chainer.training import extensions\n",
    "\n",
    "class ConvBlock(chainer.Chain):\n",
    "    \n",
    "    def __init__(self, n_ch, pool_drop=False):\n",
    "        w = chainer.initializers.HeNormal()\n",
    "        super(ConvBlock, self).__init__(\n",
    "            conv=L.Convolution2D(None, n_ch, 3, 1, 1,\n",
    "                                 nobias=True, initialW=w),\n",
    "            bn=L.BatchNormalization(n_ch)\n",
    "        )\n",
    "        \n",
    "        self.train = True\n",
    "        self.pool_drop = pool_drop\n",
    "        \n",
    "    def __call__(self, x):\n",
    "        h = F.relu(self.bn(self.conv(x)))\n",
    "        if self.pool_drop:\n",
    "            h = F.max_pooling_2d(h, 2, 2)\n",
    "            h = F.dropout(h, ratio=0.25, train=self.train)\n",
    "        return h\n",
    "    \n",
    "class LinearBlock(chainer.Chain):\n",
    "    \n",
    "    def __init__(self):\n",
    "        w = chainer.initializers.HeNormal()\n",
    "        super(LinearBlock, self).__init__(\n",
    "            fc=L.Linear(None, 1024, initialW=w))\n",
    "        self.train = True\n",
    "        \n",
    "    def __call__(self, x):\n",
    "        return F.dropout(F.relu(self.fc(x)), ratio=0.5, train=self.train)\n",
    "    \n",
    "class DeepCNN(chainer.ChainList):\n",
    "\n",
    "    def __init__(self, n_output):\n",
    "        super(DeepCNN, self).__init__(\n",
    "            ConvBlock(64),\n",
    "            ConvBlock(64, True),\n",
    "            ConvBlock(128),\n",
    "            ConvBlock(128, True),\n",
    "            ConvBlock(256),\n",
    "            ConvBlock(256),\n",
    "            ConvBlock(256),\n",
    "            ConvBlock(256, True),\n",
    "            LinearBlock(),\n",
    "            LinearBlock(),\n",
    "            L.Linear(None, n_output)\n",
    "        )\n",
    "        self._train = True\n",
    "            \n",
    "    @property\n",
    "    def train(self):\n",
    "        return self._train\n",
    "            \n",
    "    @train.setter\n",
    "    def train(self, val):\n",
    "        self._train = val\n",
    "        for c in self.children():\n",
    "            c.train = val\n",
    "    \n",
    "    def __call__(self, x):\n",
    "        for f in self.children():\n",
    "            x = f(x)\n",
    "        return x\n",
    "\n",
    "def train(model_object, batchsize=64, gpu_id=0, max_epoch=20):\n",
    "\n",
    "    # 1. Dataset\n",
    "    train, test = CIFAR10Augmented(), CIFAR10Augmented(False)\n",
    "\n",
    "    # 2. Iterator\n",
    "    train_iter = iterators.SerialIterator(train, batchsize)\n",
    "    test_iter = iterators.SerialIterator(test, batchsize, False, False)\n",
    "\n",
    "    # 3. Model\n",
    "    model = L.Classifier(model_object)\n",
    "    model.to_gpu(gpu_id)\n",
    "\n",
    "    # 4. Optimizer\n",
    "    optimizer = optimizers.Adam()\n",
    "    optimizer.setup(model)\n",
    "\n",
    "    # 5. Updater\n",
    "    updater = training.StandardUpdater(train_iter, optimizer, device=gpu_id)\n",
    "\n",
    "    # 6. Trainer\n",
    "    trainer = training.Trainer(updater, (max_epoch, 'epoch'), out='{}_cifar10augmented_result'.format(model_object.__class__.__name__))\n",
    "\n",
    "    # 7. Evaluator\n",
    "\n",
    "    class TestModeEvaluator(extensions.Evaluator):\n",
    "\n",
    "        def evaluate(self):\n",
    "            model = self.get_target('main')\n",
    "            model.train = False\n",
    "            ret = super(TestModeEvaluator, self).evaluate()\n",
    "            model.train = True\n",
    "            return ret\n",
    "\n",
    "    trainer.extend(extensions.LogReport())\n",
    "    trainer.extend(TestModeEvaluator(test_iter, model, device=gpu_id))\n",
    "    trainer.extend(extensions.PrintReport(['epoch', 'main/loss', 'main/accuracy', 'validation/main/loss', 'validation/main/accuracy', 'elapsed_time']))\n",
    "    trainer.extend(extensions.PlotReport(['main/loss', 'validation/main/loss'], x_key='epoch', file_name='loss.png'))\n",
    "    trainer.extend(extensions.PlotReport(['main/accuracy', 'validation/main/accuracy'], x_key='epoch', file_name='accuracy.png'))\n",
    "    trainer.run()\n",
    "    del trainer\n",
    "    \n",
    "    return model\n",
    "    \n",
    "model = train(DeepCNN(10), max_epoch=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "先程のData augmentationなしの場合は87%程度で頭打ちになっていた精度が、学習データにaugmentationを施すことで89%以上まで向上させられることが分かりました。2%強の改善です。\n",
    "\n",
    "最後に、ロスと精度のグラフを見てみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhMAAAEKCAYAAACoktfqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "Image(filename='DeepCNN_cifar10augmented_result/loss.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAicAAAEKCAYAAADJivWoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nReVEREREvIrKiYiIiHgVlRMRERHxKv8H8mChaU5xWVwAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='DeepCNN_cifar10augmented_result/accuracy.png')"
   ]
  }
 ],
 "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": "3.4.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}