Applications/Fashion_Recognition_(DNN_and_CNN).ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fashion Recognition Using Deep Neural Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:15:37.290046Z",
     "start_time": "2019-09-13T19:15:36.179587Z"
    }
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:16:15.720457Z",
     "start_time": "2019-09-13T19:16:15.717022Z"
    }
   },
   "outputs": [],
   "source": [
    "mnist = tf.keras.datasets.fashion_mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:17:58.009030Z",
     "start_time": "2019-09-13T19:17:48.231592Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
      "32768/29515 [=================================] - 0s 1us/step\n",
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
      "26427392/26421880 [==============================] - 7s 0us/step\n",
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
      "8192/5148 [===============================================] - 0s 0us/step\n",
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
      "4423680/4422102 [==============================] - 1s 0us/step\n"
     ]
    }
   ],
   "source": [
    "# Load the data and split as train and test\n",
    "(training_images, training_labels), (test_images, test_labels) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:18:47.800571Z",
     "start_time": "2019-09-13T19:18:47.539562Z"
    }
   },
   "outputs": [],
   "source": [
    "# Normalization\n",
    "training_images, test_images = training_images/255.0, test_images/255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:20:56.691418Z",
     "start_time": "2019-09-13T19:20:56.683442Z"
    }
   },
   "outputs": [],
   "source": [
    "# DNN model\n",
    "model = tf.keras.Sequential()\n",
    "\n",
    "model.add(tf.keras.layers.Flatten())\n",
    "model.add(tf.keras.layers.Dense(128, activation='relu'))\n",
    "model.add(tf.keras.layers.Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:21:57.979439Z",
     "start_time": "2019-09-13T19:21:57.945562Z"
    }
   },
   "outputs": [],
   "source": [
    "# Configure the model\n",
    "model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['acc'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:23:12.863746Z",
     "start_time": "2019-09-13T19:22:29.914866Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "60000/60000 [==============================] - 8s 137us/step - loss: 0.5004 - acc: 0.8241\n",
      "Epoch 2/5\n",
      "60000/60000 [==============================] - 8s 137us/step - loss: 0.3792 - acc: 0.8619\n",
      "Epoch 3/5\n",
      "60000/60000 [==============================] - 9s 148us/step - loss: 0.3388 - acc: 0.8770\n",
      "Epoch 4/5\n",
      "60000/60000 [==============================] - 9s 143us/step - loss: 0.3147 - acc: 0.8848\n",
      "Epoch 5/5\n",
      "60000/60000 [==============================] - 9s 146us/step - loss: 0.2955 - acc: 0.8912\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f35a315ceb8>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train the model\n",
    "model.fit(training_images, training_labels, epochs=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:23:43.205104Z",
     "start_time": "2019-09-13T19:23:42.757332Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 0s 44us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.36349144670963285, 0.8703]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Evaluate the model\n",
    "model.evaluate(test_images, test_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fashion Recognition Using Convolutional Neural Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:51:14.105061Z",
     "start_time": "2019-09-13T19:51:14.101277Z"
    }
   },
   "outputs": [],
   "source": [
    "mnist = tf.keras.datasets.fashion_mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:51:15.235056Z",
     "start_time": "2019-09-13T19:51:14.856346Z"
    }
   },
   "outputs": [],
   "source": [
    "# Load the data and split as train and test\n",
    "(training_images, training_labels), (test_images, test_labels) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:51:17.046708Z",
     "start_time": "2019-09-13T19:51:17.044112Z"
    }
   },
   "outputs": [],
   "source": [
    "# Reshape the data\n",
    "training_images = training_images.reshape(60000, 28, 28, 1) # 60k training images 28x28 grayscale\n",
    "test_images = test_images.reshape(10000, 28, 28, 1) # 10k test images 28x28 grayscale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:51:18.921103Z",
     "start_time": "2019-09-13T19:51:18.629652Z"
    }
   },
   "outputs": [],
   "source": [
    "# Normalization\n",
    "training_images, test_images = training_images/255.0, test_images/255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:58:20.514543Z",
     "start_time": "2019-09-13T19:58:20.419880Z"
    }
   },
   "outputs": [],
   "source": [
    "# Creating the CNN model\n",
    "model = tf.keras.models.Sequential()\n",
    "\n",
    "# First convolution layer\n",
    "model.add(tf.keras.layers.Conv2D(\n",
    "    filters=64,\n",
    "    kernel_size=(3,3),\n",
    "    activation='relu',\n",
    "    input_shape=(28,28,1)\n",
    "))\n",
    "model.add(tf.keras.layers.MaxPooling2D(\n",
    "    pool_size=(2,2)\n",
    "))\n",
    "\n",
    "# Second convolution layer\n",
    "model.add(tf.keras.layers.Conv2D(\n",
    "    filters=64,\n",
    "    kernel_size=(3,3),\n",
    "    activation='relu',\n",
    "))\n",
    "model.add(tf.keras.layers.MaxPooling2D(\n",
    "    pool_size=(2,2)\n",
    "))\n",
    "\n",
    "# DNN model\n",
    "model.add(tf.keras.layers.Flatten())\n",
    "model.add(tf.keras.layers.Dense(128, activation='relu'))\n",
    "model.add(tf.keras.layers.Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:59:35.392673Z",
     "start_time": "2019-09-13T19:59:35.349701Z"
    }
   },
   "outputs": [],
   "source": [
    "# Configure the model\n",
    "model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['acc'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T19:59:51.382898Z",
     "start_time": "2019-09-13T19:59:51.375625Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 26, 26, 64)        640       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 13, 13, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 11, 11, 64)        36928     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 5, 5, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 1600)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 128)               204928    \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 243,786\n",
      "Trainable params: 243,786\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Summary of the model\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T20:05:50.841585Z",
     "start_time": "2019-09-13T20:00:37.412681Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "60000/60000 [==============================] - 67s 1ms/step - loss: 0.4437 - acc: 0.8388\n",
      "Epoch 2/5\n",
      "60000/60000 [==============================] - 62s 1ms/step - loss: 0.3000 - acc: 0.8894\n",
      "Epoch 3/5\n",
      "60000/60000 [==============================] - 61s 1ms/step - loss: 0.2540 - acc: 0.9059\n",
      "Epoch 4/5\n",
      "60000/60000 [==============================] - 61s 1ms/step - loss: 0.2199 - acc: 0.9182\n",
      "Epoch 5/5\n",
      "60000/60000 [==============================] - 62s 1ms/step - loss: 0.1934 - acc: 0.9279\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f351b267c88>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Training the model\n",
    "model.fit(training_images, training_labels, epochs=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T20:06:40.624566Z",
     "start_time": "2019-09-13T20:06:37.224824Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 3s 339us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.27245104237794876, 0.9035]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Evaluate the model\n",
    "model.evaluate(test_images, test_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualizing the Convolutions and Pooling\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": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T22:08:25.473315Z",
     "start_time": "2019-09-13T22:08:25.466796Z"
    }
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras import models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T22:11:45.785345Z",
     "start_time": "2019-09-13T22:11:45.780315Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[9 2 1 1 6 1 4 6 5 7 4 5 7 3 4 1 2 4 8 0 2 5 7 9 1 4 6 0 9 3 8 8 3 3 8 0 7\n",
      " 5 7 9 6 1 3 7 6 7 2 1 2 2 4 4 5 8 2 2 8 4 8 0 7 7 8 5 1 1 2 3 9 8 7 0 2 6\n",
      " 2 3 1 2 8 4 1 8 5 9 5 0 3 2 0 6 5 3 6 7 1 8 0 1 4 2]\n"
     ]
    }
   ],
   "source": [
    "print(test_labels[:100])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-13T22:18:54.686424Z",
     "start_time": "2019-09-13T22:18:53.139512Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 12 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, axarr = plt.subplots(3,4)\n",
    "FIRST_IMAGE=6\n",
    "SECOND_IMAGE=10\n",
    "THIRD_IMAGE=14\n",
    "CONVOLUTION_NUMBER = 1\n",
    "\n",
    "layer_outputs = [layer.output for layer in model.layers]\n",
    "activation_model = tf.keras.models.Model(inputs = model.input, outputs = layer_outputs)\n",
    "\n",
    "for x in range(0,4):\n",
    "  f1 = activation_model.predict(test_images[FIRST_IMAGE].reshape(1, 28, 28, 1))[x]\n",
    "  axarr[0,x].imshow(f1[0, : , :, CONVOLUTION_NUMBER], cmap='inferno')\n",
    "  axarr[0,x].grid(False)\n",
    "\n",
    "  f2 = activation_model.predict(test_images[SECOND_IMAGE].reshape(1, 28, 28, 1))[x]\n",
    "  axarr[1,x].imshow(f2[0, : , :, CONVOLUTION_NUMBER], cmap='inferno')\n",
    "  axarr[1,x].grid(False)\n",
    "    \n",
    "  f3 = activation_model.predict(test_images[THIRD_IMAGE].reshape(1, 28, 28, 1))[x]\n",
    "  axarr[2,x].imshow(f3[0, : , :, CONVOLUTION_NUMBER], cmap='inferno')\n",
    "  axarr[2,x].grid(False)"
   ]
  }
 ],
 "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": "3.7.2"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}