pyimagesearch-training-your-first-cnn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Title:** Training your first Convolutional Neural Network<br/>\n",
    "**Author:** Adrian Rosebrock, [PyImageSearch.com](https://www.pyimagesearch.com/)<br/>\n",
    "\n",
    "This tutorial is meant to be an *introduction* to image classification and deep learning using the [Microsoft Data Science Virtual Machine](https://azure.microsoft.com/en-us/services/virtual-machines/data-science-virtual-machines/). We will present all code necessary to run the example, but will only briefly discuss it.\n",
    "\n",
    "For an in-depth treatment of deep learning and Convolutional Neural Networks, please refer to [*Deep Learning for Computer Vision with Python*](https://www.pyimagesearch.com/deep-learning-computer-vision-python-book/).\n",
    "\n",
    "Let's go ahead and get started."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# import the necessary packages\n",
    "from keras.models import Sequential\n",
    "from keras.layers.convolutional import Conv2D\n",
    "from keras.layers.convolutional import MaxPooling2D\n",
    "from keras.layers.core import Activation\n",
    "from keras.layers.core import Flatten\n",
    "from keras.layers.core import Dense\n",
    "from keras import backend as K\n",
    "\n",
    "class LeNet:\n",
    "\t@staticmethod\n",
    "\tdef build(width, height, depth, classes):\n",
    "\t\t# initialize the model\n",
    "\t\tmodel = Sequential()\n",
    "\t\tinputShape = (height, width, depth)\n",
    "\n",
    "\t\t# if we are using \"channels first\", update the input shape\n",
    "\t\tif K.image_data_format() == \"channels_first\":\n",
    "\t\t\tinputShape = (depth, height, width)\n",
    "            \n",
    "\t\t# first set of CONV => RELU => POOL layers\n",
    "\t\tmodel.add(Conv2D(20, (5, 5), padding=\"same\",\n",
    "\t\t\tinput_shape=inputShape))\n",
    "\t\tmodel.add(Activation(\"relu\"))\n",
    "\t\tmodel.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
    "        \n",
    "\t\t# second set of CONV => RELU => POOL layers\n",
    "\t\tmodel.add(Conv2D(50, (5, 5), padding=\"same\"))\n",
    "\t\tmodel.add(Activation(\"relu\"))\n",
    "\t\tmodel.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
    "        \n",
    "\t\t# first (and only) set of FC => RELU layers\n",
    "\t\tmodel.add(Flatten())\n",
    "\t\tmodel.add(Dense(500))\n",
    "\t\tmodel.add(Activation(\"relu\"))\n",
    "\n",
    "\t\t# softmax classifier\n",
    "\t\tmodel.add(Dense(classes))\n",
    "\t\tmodel.add(Activation(\"softmax\"))\n",
    "\n",
    "\t\t# return the constructed network architecture\n",
    "\t\treturn model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The `LeNet` class will encapsulate our implementation. The `build` method will then handle constructing the CNN. We also take special care to define the `inputShape` based on whether we are using channels-last ordering (typically used with TensorFlow) or channels-first ordering (which Theano uses).\n",
    "\n",
    "We then define our first set of `CONV => RELU => POOL` layers. Here we learn 20 convolution filters, each of which are *5x5*. We then apply a ReLU activation followed by *2x2* max-pooling to reduce the spatial dimensions.\n",
    "\n",
    "A similar process is then used for our second set of `CONV => RELU => POOL` layers, this time learning 50 filters instead of 20.\n",
    "\n",
    "Our final layer set in the LeNet implementation handles flattening out the volume into a set of fully-connected layers and output predictions.\n",
    "\n",
    "Let's move on to training our network. First, let's handle our imports and download the MNIST dataset (a collection of 70,000 28x28 grayscale images):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] accessing MNIST...\n"
     ]
    }
   ],
   "source": [
    "# import the necessary packages\n",
    "from keras.optimizers import SGD\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn import datasets\n",
    "from keras import backend as K\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "# grab the MNIST dataset (if this is your first time using this\n",
    "# dataset then the 55MB download may take a minute)\n",
    "print(\"[INFO] accessing MNIST...\")\n",
    "dataset = datasets.fetch_mldata(\"MNIST Original\")\n",
    "data = dataset.data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our goal here is to train a Convolutional Neural Network that can recognize each of the handwritten digits, 0-9.\n",
    "\n",
    "We then need to reshape the `data` matrix based on whether we are using channels-last or channels first-ordering:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# if we are using \"channels first\" ordering, then reshape the\n",
    "# design matrix such that the matrix is:\n",
    "# num_samples x depth x rows x columns\n",
    "if K.image_data_format() == \"channels_first\":\n",
    "\tdata = data.reshape(data.shape[0], 1, 28, 28)\n",
    "\n",
    "# otherwise, we are using \"channels last\" ordering, so the design\n",
    "# matrix shape should be: num_samples x rows x columns x depth\n",
    "else:\n",
    "\tdata = data.reshape(data.shape[0], 28, 28, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create the training and testing split:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# scale the input data to the range [0, 1] and perform a train/test\n",
    "# split\n",
    "(trainX, testX, trainY, testY) = train_test_split(data / 255.0,\n",
    "\tdataset.target.astype(\"int\"), test_size=0.25, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And then binarize the labels:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# convert the labels from integers to vectors\n",
    "le = LabelBinarizer()\n",
    "trainY = le.fit_transform(trainY)\n",
    "testY = le.transform(testY)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are now ready to instantiate the `LeNet` class:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] compiling model...\n"
     ]
    }
   ],
   "source": [
    "# initialize the optimizer and model\n",
    "print(\"[INFO] compiling model...\")\n",
    "opt = SGD(lr=0.01)\n",
    "model = LeNet.build(width=28, height=28, depth=1, classes=10)\n",
    "model.compile(loss=\"categorical_crossentropy\", optimizer=opt,\n",
    "\tmetrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Notice how the parameters to `LeNet.build` match the spatial dimensions of the images in the MNIST dataset.\n",
    "\n",
    "We'll be using the Stochastic Gradient Descent (SGD) optimizer for this task. Using `categorical_crossentropy` is common for multi-class classification problems.\n",
    "\n",
    "We can now train our network for 20 epochs using a batch size of 128:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] training network...\n",
      "Train on 52500 samples, validate on 17500 samples\n",
      "Epoch 1/20\n",
      "52500/52500 [==============================] - 8s 159us/step - loss: 1.0553 - acc: 0.7207 - val_loss: 0.5673 - val_acc: 0.8083\n",
      "Epoch 2/20\n",
      "52500/52500 [==============================] - 7s 124us/step - loss: 0.2861 - acc: 0.9146 - val_loss: 0.2793 - val_acc: 0.9150\n",
      "Epoch 3/20\n",
      "52500/52500 [==============================] - 7s 126us/step - loss: 0.2059 - acc: 0.9385 - val_loss: 0.3395 - val_acc: 0.8856\n",
      "Epoch 4/20\n",
      "52500/52500 [==============================] - 7s 129us/step - loss: 0.1626 - acc: 0.9519 - val_loss: 0.3899 - val_acc: 0.8765\n",
      "Epoch 5/20\n",
      "52500/52500 [==============================] - 7s 126us/step - loss: 0.1348 - acc: 0.9596 - val_loss: 0.1283 - val_acc: 0.9617\n",
      "Epoch 6/20\n",
      "52500/52500 [==============================] - 7s 127us/step - loss: 0.1148 - acc: 0.9659 - val_loss: 0.1775 - val_acc: 0.9444\n",
      "Epoch 7/20\n",
      "52500/52500 [==============================] - 7s 130us/step - loss: 0.1012 - acc: 0.9698 - val_loss: 0.1077 - val_acc: 0.9667\n",
      "Epoch 8/20\n",
      "52500/52500 [==============================] - 7s 128us/step - loss: 0.0912 - acc: 0.9720 - val_loss: 0.1718 - val_acc: 0.9415\n",
      "Epoch 9/20\n",
      "52500/52500 [==============================] - 7s 128us/step - loss: 0.0817 - acc: 0.9759 - val_loss: 0.0925 - val_acc: 0.9711\n",
      "Epoch 10/20\n",
      "52500/52500 [==============================] - 7s 128us/step - loss: 0.0755 - acc: 0.9775 - val_loss: 0.0780 - val_acc: 0.9765\n",
      "Epoch 11/20\n",
      "52500/52500 [==============================] - 7s 127us/step - loss: 0.0689 - acc: 0.9789 - val_loss: 0.4633 - val_acc: 0.8604\n",
      "Epoch 12/20\n",
      "52500/52500 [==============================] - 7s 129us/step - loss: 0.0668 - acc: 0.9794 - val_loss: 0.0732 - val_acc: 0.9783\n",
      "Epoch 13/20\n",
      "52500/52500 [==============================] - 7s 127us/step - loss: 0.0606 - acc: 0.9814 - val_loss: 0.0685 - val_acc: 0.9803\n",
      "Epoch 14/20\n",
      "52500/52500 [==============================] - 7s 129us/step - loss: 0.0568 - acc: 0.9824 - val_loss: 0.0831 - val_acc: 0.9741\n",
      "Epoch 15/20\n",
      "52500/52500 [==============================] - 7s 128us/step - loss: 0.0541 - acc: 0.9837 - val_loss: 0.0707 - val_acc: 0.9776\n",
      "Epoch 16/20\n",
      "52500/52500 [==============================] - 7s 128us/step - loss: 0.0510 - acc: 0.9844 - val_loss: 0.0595 - val_acc: 0.9829\n",
      "Epoch 17/20\n",
      "52500/52500 [==============================] - 7s 129us/step - loss: 0.0485 - acc: 0.9849 - val_loss: 0.0580 - val_acc: 0.9837\n",
      "Epoch 18/20\n",
      "52500/52500 [==============================] - 7s 131us/step - loss: 0.0463 - acc: 0.9861 - val_loss: 0.0683 - val_acc: 0.9790\n",
      "Epoch 19/20\n",
      "52500/52500 [==============================] - 7s 129us/step - loss: 0.0440 - acc: 0.9867 - val_loss: 0.0553 - val_acc: 0.9838\n",
      "Epoch 20/20\n",
      "52500/52500 [==============================] - 7s 130us/step - loss: 0.0420 - acc: 0.9871 - val_loss: 0.0569 - val_acc: 0.9821\n"
     ]
    }
   ],
   "source": [
    "# train the network\n",
    "print(\"[INFO] training network...\")\n",
    "H = model.fit(trainX, trainY, validation_data=(testX, testY),\n",
    "\tbatch_size=128, epochs=20, verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And then evaluate it on our testing set:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] evaluating network...\n",
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.99      0.99      0.99      1677\n",
      "          1       0.98      1.00      0.99      1935\n",
      "          2       0.99      0.98      0.98      1767\n",
      "          3       0.98      0.98      0.98      1766\n",
      "          4       0.98      0.99      0.99      1691\n",
      "          5       0.96      0.99      0.97      1653\n",
      "          6       1.00      0.98      0.99      1754\n",
      "          7       0.98      0.99      0.98      1846\n",
      "          8       0.99      0.96      0.97      1702\n",
      "          9       0.98      0.96      0.97      1709\n",
      "\n",
      "avg / total       0.98      0.98      0.98     17500\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# evaluate the network\n",
    "print(\"[INFO] evaluating network...\")\n",
    "predictions = model.predict(testX, batch_size=128)\n",
    "print(classification_report(testY.argmax(axis=1),\n",
    "\tpredictions.argmax(axis=1),\n",
    "\ttarget_names=[str(x) for x in le.classes_]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, let's plot the training loss and accuracy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEaCAYAAAAL7cBuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkF8vpNUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f36841d1208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the training loss and accuracy\n",
    "plt.style.use(\"ggplot\")\n",
    "plt.figure()\n",
    "plt.plot(np.arange(0, 20), H.history[\"loss\"], label=\"train_loss\")\n",
    "plt.plot(np.arange(0, 20), H.history[\"val_loss\"], label=\"val_loss\")\n",
    "plt.plot(np.arange(0, 20), H.history[\"acc\"], label=\"train_acc\")\n",
    "plt.plot(np.arange(0, 20), H.history[\"val_acc\"], label=\"val_acc\")\n",
    "plt.title(\"Training Loss and Accuracy\")\n",
    "plt.xlabel(\"Epoch #\")\n",
    "plt.ylabel(\"Loss/Accuracy\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I hope you enjoyed this tutorial!\n",
    "\n",
    "If you're interested in studying deep learning applied to computer visions in more depth, be sure to take a look at my book, [*Deep Learning for Computer Vision with Python*](https://www.pyimagesearch.com/deep-learning-computer-vision-python-book/)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}