5. predict_MNIST_with_CNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# let`s predict MNIST with CNN\n",
    "\n",
    "CNN in keras can be implemented by Conv2D layers\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\leesoojin\\Anaconda3\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D\n",
    "from keras.utils import np_utils\n",
    "from keras.callbacks import ModelCheckpoint, EarlyStopping\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "import keras.backend.tensorflow_backend as K\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "import seaborn as sns\n",
    "import os"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "load datasets mnist\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28)\n",
      "(60000,)\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import mnist\n",
    "\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "\n",
    "print(X_train.shape)\n",
    "print(y_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "reshape datasets. 28 x 28 pixel. and it`s not RGB color. set grey color"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28, 1)\n",
      "(60000, 10)\n"
     ]
    }
   ],
   "source": [
    "X_train = X_train.reshape(X_train.shape[0], 28, 28, 1).astype('float32') / 255\n",
    "X_test = X_test.reshape(X_test.shape[0], 28, 28, 1).astype('float32') / 255\n",
    "\n",
    "y_train = np_utils.to_categorical(y_train)\n",
    "y_test = np_utils.to_categorical(y_test)\n",
    "\n",
    "print(X_train.shape)\n",
    "print(y_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "with K.tf_ops.device('/device:GPU:0'):\n",
    "    model = Sequential()\n",
    "    model.add(Conv2D(32, kernel_size=(3,3), input_shape = (28,28,1), activation='relu'))\n",
    "    model.add(Conv2D(64, (3,3), activation='relu'))\n",
    "    model.add(MaxPooling2D(pool_size=2))\n",
    "    model.add(Dropout(0.25))\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128, activation='relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(y_train.shape[1], activation='softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "    \n",
    "    model_dir = './model/'\n",
    "    if not os.path.exists(model_dir):\n",
    "        os.mkdir(model_dir)\n",
    "    \n",
    "    modelName = model_dir + 'cnn_mnist.model'\n",
    "    \n",
    "    checkpointer = ModelCheckpoint(filepath=modelName, monitor='val_loss', verbose=1, save_best_only = True)\n",
    "    \n",
    "    early_stopping = EarlyStopping(monitor='val_loss', patience=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 26, 26, 32)        320       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 24, 24, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 12, 12, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 12, 12, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 9216)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 128)               1179776   \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 1,199,882\n",
      "Trainable params: 1,199,882\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 48000 samples, validate on 12000 samples\n",
      "Epoch 1/30\n",
      "48000/48000 [==============================] - 7s 143us/step - loss: 0.3120 - acc: 0.9037 - val_loss: 0.0706 - val_acc: 0.9791\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.07064, saving model to ./model/cnn_mnist.model\n",
      "Epoch 2/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0976 - acc: 0.9709 - val_loss: 0.0510 - val_acc: 0.9858\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.07064 to 0.05101, saving model to ./model/cnn_mnist.model\n",
      "Epoch 3/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0693 - acc: 0.9788 - val_loss: 0.0440 - val_acc: 0.9875\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.05101 to 0.04403, saving model to ./model/cnn_mnist.model\n",
      "Epoch 4/30\n",
      "48000/48000 [==============================] - 4s 90us/step - loss: 0.0583 - acc: 0.9821 - val_loss: 0.0429 - val_acc: 0.9875ac - ETA: 1s - loss: 0.\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.04403 to 0.04286, saving model to ./model/cnn_mnist.model\n",
      "Epoch 5/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0482 - acc: 0.9849 - val_loss: 0.0374 - val_acc: 0.9893A: 0s - loss: 0.0482 - acc: 0.\n",
      "\n",
      "Epoch 00005: val_loss improved from 0.04286 to 0.03738, saving model to ./model/cnn_mnist.model\n",
      "Epoch 6/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0425 - acc: 0.9868 - val_loss: 0.0426 - val_acc: 0.9868\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.03738\n",
      "Epoch 7/30\n",
      "48000/48000 [==============================] - 4s 90us/step - loss: 0.0379 - acc: 0.9878 - val_loss: 0.0381 - val_acc: 0.9902loss - ETA: 0s - loss: 0.0369 - ETA: 0s - loss: 0.0380 - acc: 0.98\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.03738\n",
      "Epoch 8/30\n",
      "48000/48000 [==============================] - 4s 90us/step - loss: 0.0327 - acc: 0.9891 - val_loss: 0.0338 - val_acc: 0.9905\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.03738 to 0.03376, saving model to ./model/cnn_mnist.model\n",
      "Epoch 9/30\n",
      "48000/48000 [==============================] - 4s 90us/step - loss: 0.0298 - acc: 0.9906 - val_loss: 0.0350 - val_acc: 0.99077 - - ETA: 0s - loss: 0.0\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.03376\n",
      "Epoch 10/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0271 - acc: 0.9914 - val_loss: 0.0337 - val_acc: 0.9912A: 0s - loss: 0.0267 - acc: 0.9 - ETA: 0s - loss: 0.0269 - acc: 0.99\n",
      "\n",
      "Epoch 00010: val_loss improved from 0.03376 to 0.03371, saving model to ./model/cnn_mnist.model\n",
      "Epoch 11/30\n",
      "48000/48000 [==============================] - 4s 91us/step - loss: 0.0259 - acc: 0.9916 - val_loss: 0.0352 - val_acc: 0.9905\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.03371\n",
      "Epoch 12/30\n",
      "48000/48000 [==============================] - 4s 90us/step - loss: 0.0227 - acc: 0.9926 - val_loss: 0.0351 - val_acc: 0.9904: 3s - loss: 0 - ETA: 2s - loss: 0.0235 - ETA: 1s - loss: 0.0228 - acc: 0. - ETA: 1s - loss: \n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.03371\n",
      "Epoch 13/30\n",
      "48000/48000 [==============================] - 4s 90us/step - loss: 0.0211 - acc: 0.9926 - val_loss: 0.0365 - val_acc: 0.9910TA: 1s - loss: 0.\n",
      "\n",
      "Epoch 00013: val_loss did not improve from 0.03371\n",
      "Epoch 14/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0208 - acc: 0.9930 - val_loss: 0.0345 - val_acc: 0.9912\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.03371\n",
      "Epoch 15/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0185 - acc: 0.9934 - val_loss: 0.0357 - val_acc: 0.9917\n",
      "\n",
      "Epoch 00015: val_loss did not improve from 0.03371\n",
      "Epoch 16/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0168 - acc: 0.9944 - val_loss: 0.0395 - val_acc: 0.9913 - acc\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.03371\n",
      "Epoch 17/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0165 - acc: 0.9946 - val_loss: 0.0360 - val_acc: 0.9911\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.03371\n",
      "Epoch 18/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0150 - acc: 0.9950 - val_loss: 0.0385 - val_acc: 0.99133 - \n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.03371\n",
      "Epoch 19/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0134 - acc: 0.9951 - val_loss: 0.0360 - val_acc: 0.9906\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.03371\n",
      "Epoch 20/30\n",
      "48000/48000 [==============================] - 4s 89us/step - loss: 0.0134 - acc: 0.9955 - val_loss: 0.0416 - val_acc: 0.9917\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.03371\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, y_train, epochs=30, validation_split=0.2, batch_size=200, callbacks=[early_stopping, checkpointer])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 1s 81us/step\n",
      "model accuracy : 0.9920 \n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"model accuracy : %.4f \"% model.evaluate(X_test, y_test)[1])\n",
    "\n",
    "y_vloss = history.history['val_loss']\n",
    "y_loss = history.history['loss']\n",
    "\n",
    "x_len = np.arange(len(y_loss))\n",
    "\n",
    "plt.plot(x_len, y_vloss, marker='.', c='red', label='val_set_loss')\n",
    "plt.plot(x_len, y_loss, marker='.', c='blue', label='train_set_oss')\n",
    "plt.legend()\n",
    "plt.xlabel('epochs')\n",
    "plt.ylabel('loss')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}