{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MNIST Digit Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Acquisitor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Download the mnist dataset using keras downloader, Marvin downloader can be used too."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n",
      "/home/zhang/.virtualenvs/mnist-keras-engine-env/local/lib/python2.7/site-packages/requests/__init__.py:80: RequestsDependencyWarning: urllib3 (1.23) or chardet (3.0.4) doesn't match a supported version!\n",
      "  RequestsDependencyWarning)\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "marvin_cell": "acquisitor"
   },
   "outputs": [],
   "source": [
    "\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "\n",
    "marvin_initial_dataset = {\n",
    "    \"X_train\": X_train,\n",
    "    \"y_train\": y_train,\n",
    "    \"X_test\": X_test,\n",
    "    \"y_test\": y_test\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Proportion of train/test data is 6:1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10000, 28, 28)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plot 4 images as gray scale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fea31aa6410>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.subplot(221)\n",
    "plt.imshow(X_train[0], cmap=plt.get_cmap('gray'))\n",
    "plt.subplot(222)\n",
    "plt.imshow(X_train[1], cmap=plt.get_cmap('gray'))\n",
    "plt.subplot(223)\n",
    "plt.imshow(X_train[2], cmap=plt.get_cmap('gray'))\n",
    "plt.subplot(224)\n",
    "plt.imshow(X_train[3], cmap=plt.get_cmap('gray'))\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training Preparator"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "X_train and X_test are normalized, y_train and y_test are converted from class vectors to binary class matrices."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.utils import np_utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "marvin_cell": "tpreparator"
   },
   "outputs": [],
   "source": [
    "\n",
    "X_train = marvin_initial_dataset[\"X_train\"]\n",
    "X_train = X_train.reshape(X_train.shape[0], 1, 28, 28)\n",
    "X_train = X_train.astype('float32') \n",
    "X_train /= 255\n",
    "\n",
    "X_test = marvin_initial_dataset[\"X_test\"]\n",
    "X_test = X_test.reshape(X_test.shape[0], 1, 28, 28)\n",
    "X_test = X_test.astype('float32') \n",
    "X_test /= 255\n",
    "\n",
    "nb_classes = 10\n",
    "\n",
    "y_train = np_utils.to_categorical(marvin_initial_dataset[\"y_train\"], nb_classes)\n",
    "y_test = np_utils.to_categorical(marvin_initial_dataset[\"y_test\"], nb_classes)\n",
    "\n",
    "marvin_dataset = {\n",
    "    \"X_train\": X_train,\n",
    "    \"y_train\": y_train,\n",
    "    \"X_test\": X_test,\n",
    "    \"y_test\": y_test\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Trainer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define network and train model. Convolutional Neural Networks (LeNet) used in this case."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Convolution2D, MaxPooling2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 1, 28, 28)\n"
     ]
    }
   ],
   "source": [
    "print marvin_dataset[\"X_train\"].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 10)\n"
     ]
    }
   ],
   "source": [
    "print marvin_dataset[\"y_train\"].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "marvin_cell": "trainer"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/zhang/.virtualenvs/mnist-keras-engine-env/lib/python2.7/site-packages/ipykernel_launcher.py:4: UserWarning: Update your `Conv2D` call to the Keras 2 API: `Conv2D(32, (3, 3), activation=\"relu\", input_shape=(1, 28, 28...)`\n",
      "  after removing the cwd from sys.path.\n",
      "/home/zhang/.virtualenvs/mnist-keras-engine-env/lib/python2.7/site-packages/ipykernel_launcher.py:5: UserWarning: Update your `Conv2D` call to the Keras 2 API: `Conv2D(32, (3, 3), activation=\"relu\")`\n",
      "  \"\"\"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/1\n",
      "60000/60000 [==============================] - 320s 5ms/step - loss: 0.2524 - acc: 0.9264\n"
     ]
    }
   ],
   "source": [
    "\n",
    "model = Sequential()\n",
    "model.add(Convolution2D(32, kernel_size=(3, 3), activation='relu', input_shape=(1,28,28), data_format=\"channels_first\"))\n",
    "model.add(Convolution2D(32, 3, 3, activation='relu', input_shape=(1,28,28)))\n",
    "model.add(Convolution2D(32, 3, 3, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(10, activation='softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.fit(marvin_dataset[\"X_train\"], marvin_dataset[\"y_train\"], batch_size=32, epochs=1, verbose=1)\n",
    "\n",
    "marvin_model = model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Metrics Evaluator"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Accuracy is used as Evaluation metric"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "marvin_cell": "evaluator"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 135s 2ms/step\n",
      "Accuracy is: 0.978166666667 \n"
     ]
    }
   ],
   "source": [
    "score = marvin_model.evaluate(marvin_dataset[\"X_train\"], marvin_dataset[\"y_train\"], verbose=1)\n",
    "print(\"Accuracy is: {} \".format(score[1]))\n",
    "\n",
    "marvin_metrics = {\"Accuracy\": score}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prediction Preparator"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "An image with number 4 is used as input."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_message = \"http://datawrangling.s3.amazonaws.com/sample_digit.png\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Input message for prediction should be normalized as training datasets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import cv2\n",
    "import urllib2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "marvin_cell": "ppreparator"
   },
   "outputs": [],
   "source": [
    "\n",
    "resp = urllib2.urlopen(input_message)\n",
    "img = np.asarray(bytearray(resp.read()), dtype=\"uint8\")\n",
    "\n",
    "img = cv2.imdecode(img, cv2.IMREAD_COLOR)\n",
    "img = cv2.resize(img, (28, 28))\n",
    "img = img[:,:,0]\n",
    "\n",
    "input_message = img.reshape(1, 1, 28, 28)\n",
    "input_message = input_message.astype('float32') \n",
    "input_message /= 255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predictor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Do prediction and show accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "marvin_cell": "predictor"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The image has the number [4] with 0.999991416931 accuracy\n"
     ]
    }
   ],
   "source": [
    "predicted = marvin_model.predict_classes(input_message)\n",
    "acc = marvin_model.predict(input_message)[0][predicted[0]]\n",
    "print(\"The image has the number {} with {} accuracy\".format(predicted, acc))\n",
    "\n",
    "final_prediction = predicted[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}