CIFAR10_training.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN for CIFAR-10 Photo classification\n",
    "\n",
    "**Note:\n",
    "For good practice, you should create your conda environment to run this training using the provided _requirements.yaml_ file to import the environtment with the needed dependencies.**\n",
    "\n",
    "---\n",
    "\n",
    "#\n",
    "\n",
    "First we should import the dataset form Keras and plot an example of a few images with matplotlib:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/aaronvegu/Downloads/ENTER/envs/CNN/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.23.5\n",
      "  warnings.warn(f\"A NumPy version >={np_minversion} and <{np_maxversion}\"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train: X=(50000, 32, 32, 3), Y=(50000, 1)\n",
      "Test: X=(10000, 32, 32, 3), Y=(10000, 1)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<function matplotlib.pyplot.show(close=None, block=None)>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import sys\n",
    "\n",
    "# import cifar10 dataset from Keras\n",
    "from matplotlib import pyplot\n",
    "from keras.datasets import cifar10\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import Sequential\n",
    "\n",
    "# load dataset\n",
    "(trainX, trainY), (testX, testY) = cifar10.load_data()\n",
    "\n",
    "# summarize loaded dataset\n",
    "print('Train: X=%s, Y=%s' % (trainX.shape, trainY.shape))\n",
    "print('Test: X=%s, Y=%s' % (testX.shape, testY.shape))\n",
    "\n",
    "# plot first images\n",
    "for i in range(9):\n",
    "    # define subplot\n",
    "    pyplot.subplot(330 + 1 + i)\n",
    "    # plot raw pixels\n",
    "    pyplot.imshow(trainX[i])\n",
    "\n",
    "# show figure\n",
    "pyplot.show"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we are plotting the first nine images of the dataset, with an incredibly low resolution of 32px to don't reduce the performance of the algorithm.\n",
    "\n",
    "From the prints statements we can see that there are 50,000 images in the training dataset and 10,000 in the test dataset, with 32x32 pixels resolution and 3 channels of color."
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model Evaluation\n",
    "\n",
    "CIFAR10 is already an well-defined train and test dataset that we can use, meaning that we don't really need to invest a tons of time in the preprocessing part, anyways, we are gonna use methods to ensure and measure the accuracy of our model. We can achieve that applying k-fold cross-validation with K values of 5 or 10, but that would take a lot of resources and time to complete, that's why we would choose perform evaluation in a modular way, developing a function for every piece of the process, these are:\n",
    "\n",
    "1. Load the dataset\n",
    "2. Preparation of the dataset\n",
    "3. Definition of the model\n",
    "4. Evaluation of the model\n",
    "5. Presentation of results\n",
    "\n",
    "\n",
    "#### Load Dataset\n",
    "\n",
    "We know that the images are classified in 10 classes, knowing that, we can use a one hot encoding, transforming the integer or every example into a 10 element binary vector with a 1 for the index of the class value.\n",
    "\n",
    "We can achieve this with the to_categorical() utility function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Implement this actions with load_dataset() function\n",
    "def load_dataset():\n",
    "    # load dataset\n",
    "    (trainX, trainY), (testX, testY) = cifar10.load_data()\n",
    "    # one hot encoding impl\n",
    "    trainY = to_categorical(trainY)\n",
    "    testY = to_categorical(testY)\n",
    "    return trainX, trainY, testX, testY"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Prepare pixel data\n",
    "The pixel values for each image in teh dataset are unsigned integers that goes from no color to full color (0 to 255).\n",
    "Next task is to normalize these values, we are gonna achieve that converting the unsigned integers to floats, and divide the result by the maximum value (255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The prep_pixels() function below implement these behaviors and is provided with the pixel values \n",
    "# for both the train and test datasets that will need to be scaled.\n",
    "def prep_pixels(train, test):\n",
    "    # converts from integers to floats\n",
    "    train_norm = train.astype('float32')\n",
    "    test_norm = test.astype('float32')\n",
    "    # normalize to range 0 to 1\n",
    "    train_norm = train_norm / 255.0\n",
    "    test_norm = test_norm / 255.0\n",
    "    # return normalized values\n",
    "    return train_norm, test_norm"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Define model\n",
    "We need a way to define our model, that's why we create the next function, that it will define and return the selected model, so this can be fill-in or replaced with a given model configuration that we want to evaluate. That the reason because the function is empty for the moment."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define cnn model\n",
    "def define_model():\n",
    "    model = Sequential()\n",
    "    # ...\n",
    "    return model"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Evaluate model\n",
    "This is the next step after the model has been defined\n",
    "\n",
    "To fit the model, we need the number of training epochs and batch size, for this case we will be using 100 training epochs and 64 batch size.\n",
    "\n",
    "The best approach is to use a separate validation dataset, for example splitting the train dataset into train and validation sets, but for this case, we will use the test dataset as a validation dataset for simplicity."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = define_model()\n",
    "# fit model\n",
    "history = model.fit(trainX, trainY, epochs=100, batch_size=64, validation_data=(testX, testY), verbose=0)\n",
    "# evaluate model\n",
    "_, acc = model.evaluate(testX, testY, verbose=0)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Present results\n",
    "After evaluation, we can present the results using plots.\n",
    "\n",
    "These plots give us an idea of whether the model is a good fit, overfitting or underfitting for the dataset.\n",
    "\n",
    "We will create a single figure with two subplots, one for loss and the other for accuracy. Blue lines indicate model performance on the training dataset and orange lines will indicate performance the test dataset.\n",
    "\n",
    "This function receives the data collected from the training and saves an plot image with the same name of the script and 'png' file extension."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plot of learning curves\n",
    "def summarize_diagnostics(history):\n",
    "    # loss plot\n",
    "    pyplot.subplot(211)\n",
    "    pyplot.title('Cross Entropy Loss')\n",
    "    pyplot.plot(history.history['loss'], color='blue', label='train')\n",
    "    pyplot.plot(history.history['val_loss'], color='orange', label='test')\n",
    "    # accuracy plot\n",
    "    pyplot.subplot(212)\n",
    "    pyplot.title('Classification Accuracy')\n",
    "    pyplot.plot(history.history['accuracy'], color='blue', label='train')\n",
    "    pyplot.plot(history.history['val_accuracy'], color='orange', label='test')\n",
    "    # save plot\n",
    "    filename = sys.argv[0].split('/')[-1]\n",
    "    pyplot.savefig(filename + '_plot.png')\n",
    "    pyplot.close()\n",
    "\n",
    "# optionally, we can print the final model performance on test dataset\n",
    "print('> %.3f' % (acc * 100.0))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we have all the parts of the implementation, we need a function that will drive the test harness\n",
    "\n",
    "#### Test harness function\n",
    "For this, we need call the differents functions that we already defined, so we can call this function to start the evaluation of a given model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to drive test harness for evaluating a model\n",
    "def run_test_harness():\n",
    "    # load dataset\n",
    "    trainX, trainY, testX, testY = load_dataset()\n",
    "    # prepare pixel data\n",
    "    trainX, testX = prep_pixels(trainX, testX)\n",
    "    # define model\n",
    "    model = define_model()\n",
    "    # fit model\n",
    "    history = model.fit(trainX, trainY, epochs=100, batch_size=64, validation_data=(testX, testY), verbose=0)\n",
    "    # evaluate model\n",
    "    _, acc = model.evaluate(testX, testY, verbose=0)\n",
    "    print('> %.3f' % (acc * 100.0))\n",
    "    # plot learning curves\n",
    "    summarize_diagnostics(history)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Complete Example\n",
    "We had been developing the training process by portions of code, below we can see the full example that we already developed:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# test harness for evaluating models on the cifar10 dataset\n",
    "import sys\n",
    "from matplotlib import pyplot\n",
    "from keras.datasets import cifar10\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D\n",
    "from keras.layers import MaxPooling2D\n",
    "from keras.layers import Dense\n",
    "from keras.layers import Flatten\n",
    "from keras.optimizers import SGD\n",
    " \n",
    "# load train and test dataset\n",
    "def load_dataset():\n",
    " # load dataset\n",
    " (trainX, trainY), (testX, testY) = cifar10.load_data()\n",
    " # one hot encode target values\n",
    " trainY = to_categorical(trainY)\n",
    " testY = to_categorical(testY)\n",
    " return trainX, trainY, testX, testY\n",
    " \n",
    "# scale pixels\n",
    "def prep_pixels(train, test):\n",
    " # convert from integers to floats\n",
    " train_norm = train.astype('float32')\n",
    " test_norm = test.astype('float32')\n",
    " # normalize to range 0-1\n",
    " train_norm = train_norm / 255.0\n",
    " test_norm = test_norm / 255.0\n",
    " # return normalized images\n",
    " return train_norm, test_norm\n",
    " \n",
    "# define cnn model\n",
    "def define_model():\n",
    " model = Sequential()\n",
    " # ...\n",
    " return model\n",
    " \n",
    "# plot diagnostic learning curves\n",
    "def summarize_diagnostics(history):\n",
    " # plot loss\n",
    " pyplot.subplot(211)\n",
    " pyplot.title('Cross Entropy Loss')\n",
    " pyplot.plot(history.history['loss'], color='blue', label='train')\n",
    " pyplot.plot(history.history['val_loss'], color='orange', label='test')\n",
    " # plot accuracy\n",
    " pyplot.subplot(212)\n",
    " pyplot.title('Classification Accuracy')\n",
    " pyplot.plot(history.history['accuracy'], color='blue', label='train')\n",
    " pyplot.plot(history.history['val_accuracy'], color='orange', label='test')\n",
    " # save plot to file\n",
    " filename = sys.argv[0].split('/')[-1]\n",
    " pyplot.savefig(filename + '_plot.png')\n",
    " pyplot.close()\n",
    " \n",
    "# run the test harness for evaluating a model\n",
    "def run_test_harness():\n",
    " # load dataset\n",
    " trainX, trainY, testX, testY = load_dataset()\n",
    " # prepare pixel data\n",
    " trainX, testX = prep_pixels(trainX, testX)\n",
    " # define model\n",
    " model = define_model()\n",
    " # fit model\n",
    " history = model.fit(trainX, trainY, epochs=100, batch_size=64, validation_data=(testX, testY), verbose=0)\n",
    " # evaluate model\n",
    " _, acc = model.evaluate(testX, testY, verbose=0)\n",
    " print('> %.3f' % (acc * 100.0))\n",
    " # learning curves\n",
    " summarize_diagnostics(history)\n",
    " \n",
    "# entry point, run the test harness\n",
    "run_test_harness()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This example can't be run because there is no model defined, for that, is better is we first define and evaluate a baseline model.\n",
    "\n",
    "#### Defining a baseline model\n",
    "A baseline model will leet us establish a minimum model performance to which compare all other models that we can try and a model architecture to follow as a basis of improvement.\n",
    "\n",
    "We choose the architectural principles of the VGG models, because is a model that has proved a good performance and the modular structure of the architecture is easy to understand and follow.\n",
    "\n",
    "The architecture is formed by the stacking of 3x3 filters followed by a max pooling layer, together form a block, and these blocks can be repeated, givings us the modular architecture that we mentioned earlier. We can repeat the blocks increasing the depth of the networks such as 32, 64, 128, 256 and so on.\n",
    "\n",
    "#### Example of a 1-block VGG style architecture\n",
    "Below we see the definition of our model architecture with VGG using only 1 block"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# model definition with 1 block\n",
    "def define_model():\n",
    " model = Sequential()\n",
    " model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same', input_shape=(32, 32, 3)))\n",
    " model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To interpret the results, we are gonna need an output layer with 10 nodes for the 10 classes in our dataset, and use the softmax activation function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.add(MaxPooling2D((2, 2)))\n",
    "model.add(Flatten())"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For optimization, we will be using SGD or Stochastic Gradient Descendent with a learning rate of 0.001 and a large momentum of 0.9 for starting points. The model will optimize the categorical cross entropy loss function required for multi-class classification and will monitor classification accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compile model\n",
    "opt = SGD(lr=0.001, momentum=0.9)\n",
    "model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With that, we can create separate functions for every number of block that we want to test, for simplicity we will create a model definition with 3-blocks and get the results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/aaronvegu/Downloads/ENTER/envs/CNN/lib/python3.10/site-packages/keras/optimizers/optimizer_v2/gradient_descent.py:111: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
      "  super().__init__(name, **kwargs)\n",
      "2023-03-24 14:12:32.959089: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# test harness for evaluating models on the cifar10 dataset\n",
    "import sys\n",
    "from matplotlib import pyplot\n",
    "from keras.datasets import cifar10\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D\n",
    "from keras.layers import MaxPooling2D\n",
    "from keras.layers import Dense\n",
    "from keras.layers import Flatten\n",
    "from keras.optimizers import SGD\n",
    " \n",
    "# load train and test dataset\n",
    "def load_dataset():\n",
    " # load dataset\n",
    " (trainX, trainY), (testX, testY) = cifar10.load_data()\n",
    " # one hot encode target values\n",
    " trainY = to_categorical(trainY)\n",
    " testY = to_categorical(testY)\n",
    " return trainX, trainY, testX, testY\n",
    " \n",
    "# scale pixels\n",
    "def prep_pixels(train, test):\n",
    " # convert from integers to floats\n",
    " train_norm = train.astype('float32')\n",
    " test_norm = test.astype('float32')\n",
    " # normalize to range 0-1\n",
    " train_norm = train_norm / 255.0\n",
    " test_norm = test_norm / 255.0\n",
    " # return normalized images\n",
    " return train_norm, test_norm\n",
    " \n",
    "# define cnn model with 3-blocks VGG\n",
    "def define_model():\n",
    " model = Sequential()\n",
    " model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same', input_shape=(32, 32, 3)))\n",
    " model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
    " model.add(MaxPooling2D((2, 2)))\n",
    " model.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
    " model.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
    " model.add(MaxPooling2D((2, 2)))\n",
    " model.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
    " model.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
    " model.add(MaxPooling2D((2, 2)))\n",
    " model.add(Flatten())\n",
    " model.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))\n",
    " model.add(Dense(10, activation='softmax'))\n",
    " # compile model\n",
    " opt = SGD(lr=0.001, momentum=0.9)\n",
    " model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])\n",
    " return model\n",
    " \n",
    "# plot diagnostic learning curves\n",
    "def summarize_diagnostics(history):\n",
    " # plot loss\n",
    " pyplot.subplot(211)\n",
    " pyplot.title('Cross Entropy Loss')\n",
    " pyplot.plot(history.history['loss'], color='blue', label='train')\n",
    " pyplot.plot(history.history['val_loss'], color='orange', label='test')\n",
    " # plot accuracy\n",
    " pyplot.subplot(212)\n",
    " pyplot.title('Classification Accuracy')\n",
    " pyplot.plot(history.history['accuracy'], color='blue', label='train')\n",
    " pyplot.plot(history.history['val_accuracy'], color='orange', label='test')\n",
    " # save plot to file\n",
    " filename = sys.argv[0].split('/')[-1]\n",
    " pyplot.savefig(filename + '_plot.png')\n",
    " pyplot.close()\n",
    " \n",
    "# run the test harness for evaluating a model\n",
    "def run_test_harness():\n",
    " # load dataset\n",
    " trainX, trainY, testX, testY = load_dataset()\n",
    " # prepare pixel data\n",
    " trainX, testX = prep_pixels(trainX, testX)\n",
    " # define model\n",
    " model = define_model()\n",
    " # fit model\n",
    " history = model.fit(trainX, trainY, epochs=100, batch_size=64, validation_data=(testX, testY), verbose=0)\n",
    " # evaluate model\n",
    " _, acc = model.evaluate(testX, testY, verbose=0)\n",
    " print('> %.3f' % (acc * 100.0))\n",
    " # learning curves\n",
    " summarize_diagnostics(history)\n",
    " \n",
    "# entry point, run the test harness\n",
    "run_test_harness()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.10.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}