fashion-mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fashion-MNIST\n",
    "\n",
    "Dataset of 60,000 28x28 grayscale images of 10 fashion categories, along with a test set of 10,000 images. This dataset can be used as a drop-in replacement for MNIST. The class labels are:\n",
    "\n",
    "| Label | Description |\n",
    "|---|---|\n",
    "| 0 | T-shirt/top |\n",
    "| 1 | Trouser |\n",
    "| 2 | Pullover |\n",
    "| 3 | Dress |\n",
    "| 4 | Coat |\n",
    "| 5 | Sandal |\n",
    "| 6 | Shirt |\n",
    "| 7 | Sneaker |\n",
    "| 8 | Bag |\n",
    "| 9 | Ankle boot |"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import classification_report\n",
    "import keras\n",
    "from keras.datasets import fashion_mnist\n",
    "from keras.utils import to_categorical\n",
    "from keras import Sequential\n",
    "from keras.layers import MaxPooling2D\n",
    "from keras.layers import Conv2D, Dense, Flatten, Dropout, Activation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Load the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the dataset.\n",
    "(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()\n",
    "\n",
    "# Set the number of classes.\n",
    "num_classes = np.max(y_train) + 1\n",
    "\n",
    "# Set the labels.\n",
    "labels = [\"T-shirt/top\", \"Trouser\", \"Pullover\", \"Dress\", \n",
    "          \"Coat\", \"Sandal\", \"Shirt\", \"Sneaker\", \"Bag\", \"Ankle boot\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Display an image for each class.\n",
    "fig = plt.figure(figsize=(15,5))\n",
    "for i in range(num_classes):\n",
    "    ax = fig.add_subplot(2, 5, 1 + i, xticks=[], yticks=[])\n",
    "    idx = np.where(y_train[:] == i)[0]\n",
    "    features_idx = x_train[idx,::]\n",
    "    img_num = np.random.randint(features_idx.shape[0])\n",
    "    im = features_idx[img_num]\n",
    "    plt.imshow(im, cmap=\"gray\")\n",
    "    plt.title(\"{}\".format(labels[i]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Handle the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_tr size = (60000, 28, 28, 1)\n",
      "x_te size = (10000, 28, 28, 1)\n"
     ]
    }
   ],
   "source": [
    "# Reshape as channel first.\n",
    "size_tr = x_train.shape\n",
    "size_te = x_test.shape\n",
    "x_tr = x_train.reshape(size_tr[0], size_tr[1], size_tr[2], 1)\n",
    "x_te = x_test.reshape(size_te[0], size_te[1], size_te[2], 1)\n",
    "\n",
    "# Process the images.\n",
    "x_tr = x_tr / 255\n",
    "x_te = x_te / 255\n",
    "\n",
    "# Display the images size.\n",
    "print(\"x_tr size = {}\".format(x_tr.shape))\n",
    "print(\"x_te size = {}\".format(x_te.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y_tr size = (60000, 10)\n",
      "y_te size = (10000, 10)\n"
     ]
    }
   ],
   "source": [
    "# Process the classes.\n",
    "y_tr = to_categorical(y_train, num_classes)\n",
    "y_te = to_categorical(y_test, num_classes)\n",
    "\n",
    "# Display the classes size.\n",
    "print(\"y_tr size = {}\".format(y_tr.shape))\n",
    "print(\"y_te size = {}\".format(y_te.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Train the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Variable *= will be deprecated. Use variable.assign_mul if you want assignment to the variable value or 'x = x * y' if you want a new python Tensor object.\n",
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 50s 839us/step - loss: 0.5877 - acc: 0.7869 - val_loss: 0.4119 - val_acc: 0.8557\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 50s 828us/step - loss: 0.3690 - acc: 0.8674 - val_loss: 0.3316 - val_acc: 0.8799\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 47s 785us/step - loss: 0.3143 - acc: 0.8877 - val_loss: 0.2975 - val_acc: 0.8919\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 47s 776us/step - loss: 0.2835 - acc: 0.8978 - val_loss: 0.2812 - val_acc: 0.8985\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 47s 776us/step - loss: 0.2604 - acc: 0.9052 - val_loss: 0.2704 - val_acc: 0.9020\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 45s 758us/step - loss: 0.2433 - acc: 0.9105 - val_loss: 0.2597 - val_acc: 0.9037\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 45s 757us/step - loss: 0.2297 - acc: 0.9159 - val_loss: 0.2529 - val_acc: 0.9085\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 50s 839us/step - loss: 0.2126 - acc: 0.9224 - val_loss: 0.2576 - val_acc: 0.9053\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 47s 784us/step - loss: 0.1987 - acc: 0.9259 - val_loss: 0.2470 - val_acc: 0.9109\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 48s 795us/step - loss: 0.1897 - acc: 0.9296 - val_loss: 0.2429 - val_acc: 0.9135\n"
     ]
    }
   ],
   "source": [
    "# Initialise the the model.\n",
    "model = Sequential()\n",
    "\n",
    "# First convolutional layer.\n",
    "model.add(Conv2D(32, kernel_size=(5, 5), strides=(1, 1),\n",
    "                 activation=\"relu\",\n",
    "                 data_format=\"channels_last\",\n",
    "                 input_shape=(28, 28, 1)))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
    "\n",
    "# Second convolutional layer.\n",
    "model.add(Conv2D(64, (5, 5), activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "# Dense layer.\n",
    "model.add(Flatten())\n",
    "model.add(Dense(256, activation='relu'))\n",
    "model.add(Dropout(0.50))\n",
    "model.add(Dense(num_classes, activation='softmax'))\n",
    "\n",
    "# Set the optimizer.\n",
    "optimizer = keras.optimizers.Adam(lr=0.001, decay=1e-6)\n",
    "\n",
    "# Compile the model.\n",
    "model.compile(loss=\"categorical_crossentropy\",\n",
    "              optimizer=optimizer,\n",
    "              metrics=[\"accuracy\"])\n",
    "\n",
    "# Parameter fitting.\n",
    "fit_params = {\n",
    "    \"batch_size\": 128,\n",
    "    \"epochs\": 10,\n",
    "    \"verbose\": 1,\n",
    "    \"shuffle\": True,\n",
    "    \"validation_data\": (x_te, y_te)\n",
    "}\n",
    "\n",
    "# Fit the model.\n",
    "history = model.fit(x_tr, y_tr, **fit_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualise the training resuls.\n",
    "plt.figure(figsize=(15,5))\n",
    "plt.subplot(121)\n",
    "plt.plot(history.history[\"loss\"], color=\"b\", label=\"tr\")\n",
    "plt.plot(history.history[\"val_loss\"], color=\"r\", label=\"te\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.xlabel(\"epochs\")\n",
    "plt.grid()\n",
    "plt.subplot(122)\n",
    "plt.plot(history.history[\"acc\"], color=\"b\", label=\"tr\")\n",
    "plt.plot(history.history[\"val_acc\"], color=\"r\", label=\"te\")\n",
    "plt.ylabel(\"acc\")\n",
    "plt.xlabel(\"epochs\")\n",
    "plt.grid()\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the model.\n",
    "model.save(\"models/fashion_mnist.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Evaluate the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set the model filepath.\n",
    "filepath = \"models/fashion_mnist.h5\"\n",
    "\n",
    "# Load the model.\n",
    "model = keras.models.load_model(filepath)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "T-shirt/top     0.8617    0.8720    0.8668      1000\n",
      "    Trouser     0.9949    0.9790    0.9869      1000\n",
      "   Pullover     0.8871    0.8490    0.8677      1000\n",
      "      Dress     0.9359    0.8760    0.9050      1000\n",
      "       Coat     0.8229    0.9110    0.8647      1000\n",
      "     Sandal     0.9802    0.9880    0.9841      1000\n",
      "      Shirt     0.7480    0.7480    0.7480      1000\n",
      "    Sneaker     0.9451    0.9810    0.9627      1000\n",
      "        Bag     0.9829    0.9800    0.9815      1000\n",
      " Ankle boot     0.9896    0.9510    0.9699      1000\n",
      "\n",
      "avg / total     0.9148    0.9135    0.9137     10000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Model test inference.\n",
    "y_pred = model.predict(x_te)\n",
    "y_pred = np.eye(num_classes)[np.argmax(y_pred, axis=1).reshape(-1)]\n",
    "\n",
    "# Display the classification report.\n",
    "print(classification_report(y_te, y_pred, target_names=labels, digits=4))"
   ]
  },
  {
   "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}