CNN on Fashoin Mnist dataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.2.0-rc4'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix, accuracy_score, precision_score, classification_report\n",
    "def metrics(y_test, y_pred):\n",
    "    y_pred = y_pred.argmax(axis=1)\n",
    "    print(y_pred)\n",
    "    print(\"Accuracy Score : \", accuracy_score(y_test, y_pred))\n",
    "    print(\"Precision Score : \", precision_score(y_test, y_pred, average=\"weighted\"))\n",
    "    print(\"Classification Report : \\n\", classification_report(y_test, y_pred))\n",
    "    print(\"Confusion Matrix : \\n\", confusion_matrix(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "## Load the data\n",
    "\n",
    "<hr>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X train shape : (60000, 28, 28)\n",
      "y train shape : (60000,)\n",
      "X test shape : (10000, 28, 28)\n",
      "y test shape : (10000,)\n"
     ]
    }
   ],
   "source": [
    "fashion_mnist = tf.keras.datasets.fashion_mnist\n",
    "\n",
    "(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()\n",
    "\n",
    "# Always a good idea to look at shape before getting started.\n",
    "print(f\"X train shape : {X_train.shape}\")\n",
    "print(f\"y train shape : {y_train.shape}\")\n",
    "print(f\"X test shape : {X_test.shape}\")\n",
    "print(f\"y test shape : {y_test.shape}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Now, if we see training set contains 60000 samples. \n",
    "##### Testing set has 10000. \n",
    "##### As shape of each image is 28 * 28 pixels (picture element) --> can be interpreted as grayscale as no color layer is specified.\n",
    "\n",
    "##### In grayscale 255 --> white , 0 --> black and any other value --> shade of grey."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "# All of the above are numpy arrays --> run much faster than python lists.\n",
    "# numpy is implemented in C.\n",
    "\n",
    "# tensorflow is named after \"tensor\" which is a generalised numpy array.\n",
    "\n",
    "print(type(X_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New X_train shape : (60000, 28, 28, 1)\n",
      "New X_test shape : (10000, 28, 28, 1)\n"
     ]
    }
   ],
   "source": [
    "# Normalising values as gradient descent works better on small scale data.\n",
    "\n",
    "X_train , X_test = X_train/255, X_test/255\n",
    "\n",
    "# Converting input suitable to CNN architecture.\n",
    "\n",
    "X_train = np.reshape(X_train,newshape=(X_train.shape+(1,)))\n",
    "\n",
    "print(f\"New X_train shape : {X_train.shape}\")\n",
    "\n",
    "X_test = np.reshape(X_test, newshape=(X_test.shape[0], X_test.shape[1], X_test.shape[2], 1))\n",
    "\n",
    "print(f\"New X_test shape : {X_test.shape}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Now, input is suitable for Convolutional neural network.\n",
    "##### Also, it is properly scaled for gradient descent"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New shape of y_train (60000,)\n",
      "New shape of y_test (10000,)\n"
     ]
    }
   ],
   "source": [
    "# Trying to make output into proper format\n",
    "\n",
    "y_train, y_test = y_train.flatten(), y_test.flatten()\n",
    "\n",
    "print(f\"New shape of y_train {y_train.shape}\")\n",
    "print(f\"New shape of y_test {y_test.shape}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Label mapping\n",
    "labels = '''T-shirt/top\n",
    "Trouser\n",
    "Pullover\n",
    "Dress\n",
    "Coat\n",
    "Sandal\n",
    "Shirt\n",
    "Sneaker\n",
    "Bag\n",
    "Ankle boot'''.split(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No. of classes : 10\n"
     ]
    }
   ],
   "source": [
    "# Getting number of classes\n",
    "K = len(set(y_train))\n",
    "print(f\"No. of classes : {K}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "## Building the model\n",
    "\n",
    "<hr>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(28, 28, 1)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Buliding model using keras functional API.\n",
    "\n",
    "from tensorflow.keras.layers import Input, Conv2D, Flatten, Dense, Dropout\n",
    "from tensorflow.keras.models import Model\n",
    "\n",
    "\n",
    "input_shape = X_train[0].shape\n",
    "feature_map = [32, 64, 128]\n",
    "dense_size = [512]\n",
    "filter_size= (3,3)\n",
    "\n",
    "i = Input(shape=input_shape)\n",
    "x = Conv2D(filters=feature_map[0], kernel_size=filter_size, activation=\"relu\", strides=2)(i)\n",
    "x = Conv2D(filters=feature_map[1], kernel_size=filter_size, activation=\"relu\", strides=2)(x)\n",
    "x = Conv2D(filters=feature_map[2], kernel_size=filter_size, activation=\"relu\", strides=2)(x)\n",
    "\n",
    "# Dense layer takes flattened input\n",
    "x = Flatten()(x)\n",
    "\n",
    "# Building dense/ fully connected layers from here\n",
    "x = Dense(dense_size[0], activation=\"relu\")(x)\n",
    "x = Dense(K, activation=\"softmax\")(x)\n",
    "\n",
    "model = Model(i, x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.4884 - accuracy: 0.8173 - val_loss: 0.3959 - val_accuracy: 0.8519\n",
      "Epoch 2/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.3347 - accuracy: 0.8741 - val_loss: 0.3377 - val_accuracy: 0.8759\n",
      "Epoch 3/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.2832 - accuracy: 0.8934 - val_loss: 0.3117 - val_accuracy: 0.8805\n",
      "Epoch 4/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.2470 - accuracy: 0.9059 - val_loss: 0.3233 - val_accuracy: 0.8846\n",
      "Epoch 5/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.2170 - accuracy: 0.9178 - val_loss: 0.3040 - val_accuracy: 0.8931\n",
      "Epoch 6/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1902 - accuracy: 0.9274 - val_loss: 0.3097 - val_accuracy: 0.8965\n",
      "Epoch 7/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1662 - accuracy: 0.9360 - val_loss: 0.3194 - val_accuracy: 0.8955\n",
      "Epoch 8/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1452 - accuracy: 0.9448 - val_loss: 0.3509 - val_accuracy: 0.8955\n",
      "Epoch 9/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1260 - accuracy: 0.9518 - val_loss: 0.3721 - val_accuracy: 0.8946\n",
      "Epoch 10/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1088 - accuracy: 0.9584 - val_loss: 0.4198 - val_accuracy: 0.8953\n",
      "Epoch 11/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0973 - accuracy: 0.9626 - val_loss: 0.4107 - val_accuracy: 0.8974\n",
      "Epoch 12/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0832 - accuracy: 0.9682 - val_loss: 0.4540 - val_accuracy: 0.8973\n",
      "Epoch 13/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0753 - accuracy: 0.9714 - val_loss: 0.5108 - val_accuracy: 0.8940\n",
      "Epoch 14/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0671 - accuracy: 0.9747 - val_loss: 0.4766 - val_accuracy: 0.8994\n",
      "Epoch 15/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0600 - accuracy: 0.9767 - val_loss: 0.5489 - val_accuracy: 0.8910\n",
      "Epoch 16/20\n",
      "1875/1875 [==============================] - 13s 7ms/step - loss: 0.0538 - accuracy: 0.9800 - val_loss: 0.5832 - val_accuracy: 0.8953\n",
      "Epoch 17/20\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.0535 - accuracy: 0.9805 - val_loss: 0.5932 - val_accuracy: 0.8947\n",
      "Epoch 18/20\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.0490 - accuracy: 0.9815 - val_loss: 0.6232 - val_accuracy: 0.8989\n",
      "Epoch 19/20\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0434 - accuracy: 0.9841 - val_loss: 0.6660 - val_accuracy: 0.8958\n",
      "Epoch 20/20\n",
      "1875/1875 [==============================] - 10s 6ms/step - loss: 0.0445 - accuracy: 0.9839 - val_loss: 0.6537 - val_accuracy: 0.8991\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer=\"adam\", loss=\"sparse_categorical_crossentropy\", metrics=[\"accuracy\"])\n",
    "\n",
    "r = model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Early Stopping --> Stop when loss of test dataset increases"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "## Visualsing results of model\n",
    "\n",
    "<hr>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r.history.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'No. of epochs')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(r.history[\"accuracy\"], label=\"Accuracy\")\n",
    "plt.plot(r.history[\"val_accuracy\"], label=\"Validation Acc.\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"No. of epochs\")\n",
    "plt.ylabel(\"Accuracy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x14ee31be0>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(r.history[\"loss\"], label=\"Loss\")\n",
    "plt.plot(r.history[\"val_loss\"], label=\"Validation Loss\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### If we observe above model is clearly overfitting. \n",
    "#### We can regularize by providing it with Dropout or some kernel l2 regularizer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "i = Input(shape=(X_train[0].shape))\n",
    "x = Conv2D(32, (3,3), strides=2, activation=\"relu\")(i)\n",
    "x = Conv2D(64, (3,3), strides=2, activation=\"relu\")(x)\n",
    "x = Conv2D(128, (3,3), strides=2, activation=\"relu\")(x)\n",
    "\n",
    "x = Flatten()(x)\n",
    "\n",
    "x = Dropout(0.2)(x)\n",
    "x = Dense(128, activation=\"relu\")(x)\n",
    "x = Dropout(0.2)(x)\n",
    "x = Dense(K, activation=\"softmax\")(x)\n",
    "\n",
    "model = Model(i,x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "1875/1875 [==============================] - 8s 4ms/step - loss: 0.5525 - accuracy: 0.7971 - val_loss: 0.4237 - val_accuracy: 0.8425\n",
      "Epoch 2/5\n",
      "1875/1875 [==============================] - 9s 5ms/step - loss: 0.3791 - accuracy: 0.8582 - val_loss: 0.3602 - val_accuracy: 0.8695\n",
      "Epoch 3/5\n",
      "1875/1875 [==============================] - 8s 4ms/step - loss: 0.3287 - accuracy: 0.8768 - val_loss: 0.3345 - val_accuracy: 0.8740\n",
      "Epoch 4/5\n",
      "1875/1875 [==============================] - 9s 5ms/step - loss: 0.2977 - accuracy: 0.8883 - val_loss: 0.3193 - val_accuracy: 0.8849\n",
      "Epoch 5/5\n",
      "1875/1875 [==============================] - 8s 4ms/step - loss: 0.2704 - accuracy: 0.8995 - val_loss: 0.3059 - val_accuracy: 0.8879\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer=\"adam\", loss=\"sparse_categorical_crossentropy\", metrics=[\"accuracy\"])\n",
    "\n",
    "r = model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Accuracy')"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(r.history[\"accuracy\"], label=\"Accuracy\")\n",
    "plt.plot(r.history[\"val_accuracy\"], label=\"Validation Acc.\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"No. of epochs\")\n",
    "plt.ylabel(\"Accuracy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x13806c400>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(r.history[\"loss\"], label=\"Loss\")\n",
    "plt.plot(r.history[\"val_loss\"], label=\"Validation Loss\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[9 2 1 ... 8 1 5]\n",
      "Accuracy Score :  0.8879\n",
      "Precision Score :  0.8889139490184038\n",
      "Classification Report :                precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.86      0.85      1000\n",
      "           1       0.99      0.97      0.98      1000\n",
      "           2       0.81      0.83      0.82      1000\n",
      "           3       0.89      0.90      0.90      1000\n",
      "           4       0.84      0.77      0.81      1000\n",
      "           5       0.99      0.96      0.97      1000\n",
      "           6       0.66      0.69      0.67      1000\n",
      "           7       0.95      0.96      0.95      1000\n",
      "           8       0.97      0.97      0.97      1000\n",
      "           9       0.95      0.97      0.96      1000\n",
      "\n",
      "    accuracy                           0.89     10000\n",
      "   macro avg       0.89      0.89      0.89     10000\n",
      "weighted avg       0.89      0.89      0.89     10000\n",
      "\n",
      "Confusion Matrix : \n",
      " [[858   1  24  14   1   1  96   0   5   0]\n",
      " [  1 968   1  20   2   0   7   0   1   0]\n",
      " [ 12   2 833  10  47   0  93   0   3   0]\n",
      " [ 23   3  17 900  24   0  31   0   2   0]\n",
      " [  0   1  74  36 774   0 113   0   2   0]\n",
      " [  0   0   0   2   0 960   0  24   2  12]\n",
      " [127   0  80  25  65   0 685   0  18   0]\n",
      " [  0   0   0   0   0   5   0 958   0  37]\n",
      " [  1   0   4   4   4   4   6   3 974   0]\n",
      " [  0   0   0   0   0   3   1  27   0 969]]\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(X_test)\n",
    "metrics(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<hr>\n",
    "\n",
    "# Misclassified points\n",
    "\n",
    "<hr>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([  21,   23,   25, ..., 9972, 9977, 9991])"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "misclassified_index = np.where(y_pred.argmax(axis=1) != y_test)\n",
    "X_test = X_test * 255\n",
    "misclassified_index[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = np.random.choice(misclassified_index[0])\n",
    "plt.imshow(X_test[i].reshape(28,28), cmap='gray')\n",
    "plt.title(\"True label: %s Predicted: %s\" % (labels[y_test[i]], labels[y_pred.argmax(axis=1)[i]]));"
   ]
  }
 ],
 "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": "3.8.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}