Code_CNN_Digits_Recognition.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "InzAkXaJ9DCl"
   },
   "source": [
    "#**Librairies Importation**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "redW7TwE8dgn"
   },
   "outputs": [],
   "source": [
    "#Data Visualization\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "#import Deep Learning librairies\n",
    "\n",
    "import tensorflow as tf\n",
    "import keras\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.wrappers.scikit_learn import KerasClassifier\n",
    "from keras.utils.np_utils import to_categorical # convert to one-hot-encoding\n",
    "from tensorflow.keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D\n",
    "from tensorflow.keras.optimizers import RMSprop\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.callbacks import ReduceLROnPlateau\n",
    "\n",
    "# Sampling\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "mXfekdEZ83Uk"
   },
   "source": [
    "# **DATA IMPORTATION & SHAPING**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6N6b_FuP8zRT"
   },
   "outputs": [],
   "source": [
    "train = pd.read_csv('train.csv').drop('label',axis =1)\n",
    "label = pd.read_csv('train.csv')['label']\n",
    "test = pd.read_csv('test.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "6NEvdQRN66N2"
   },
   "source": [
    "Let's check the distribution of the labels to see if we need to make a smart sampling(upsampling, downsampling, smote, etc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 202
    },
    "colab_type": "code",
    "id": "XesVLHgs63ib",
    "outputId": "e0926a3a-a3f6-40ea-c55e-4a33ff41c041"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    4684\n",
       "7    4401\n",
       "3    4351\n",
       "9    4188\n",
       "2    4177\n",
       "6    4137\n",
       "0    4132\n",
       "4    4072\n",
       "8    4063\n",
       "5    3795\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 19,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 332
    },
    "colab_type": "code",
    "id": "YWgmd6KQGnl8",
    "outputId": "c249a8aa-456f-4df9-bdfe-52616216d510"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([4132., 4684., 4177., 4351., 4072., 3795., 4137., 4401., 4063.,\n",
       "        4188.]),\n",
       " array([0. , 0.9, 1.8, 2.7, 3.6, 4.5, 5.4, 6.3, 7.2, 8.1, 9. ]),\n",
       " <a list of 10 Patch objects>)"
      ]
     },
     "execution_count": 29,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD4CAYAAAAAczaOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(label)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "nc4XTaJh7F7B"
   },
   "source": [
    "As we can see, the data is well distributed. Hence, we can work on the label dataset with no specific sampling."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TQvyvQHw9K2Z"
   },
   "outputs": [],
   "source": [
    "#Normalization\n",
    "\n",
    "train = train/255\n",
    "test = test/255\n",
    "\n",
    "#Reshape to create images 28x28x1\n",
    "\n",
    "train = train.values.reshape(-1,28,28,1)\n",
    "test = test.values.reshape(-1,28,28,1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "V4prRpZs7q3t"
   },
   "source": [
    "First question to ask: Which model do I use?\n",
    "It is obvious that Deep Learning is more adapted and research papers and articles on the net suggests that CNN is a very common model for such topic. Then, I will implement a **CNN on Tensorflow 2.0**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "F23-dcGB9LNj"
   },
   "source": [
    "To have a concise notebook, I have chosen this time to implement what I consider being the most important steps. Here I will discuss details I have worked on but didn't choose to put them in the Notebook:\n",
    "\n",
    "* For the optimizer, I have chosen **Adam optimizer** as I have tried **RMSprop** (*lower accuracy by ~ 1%*) & **AdaDelta** (*very slow to converge*)\n",
    "\n",
    "* In researchs I have made, the pool size = (2,2) is common and we don't need to change it\n",
    "\n",
    "* To avoid overfitting, I have chosen to work for the first time to **Dropout**\n",
    "\n",
    "* For the cross validation, I choose a **5-fold (80%/20%)**\n",
    "\n",
    "* The number of Conv2D, I have tried one, two or three. One Conv2D was not efficient and three Conv2D were less efficient than two Conv2D. Then, **[2 Conv2D -> MaxPool2D -> Droupout]** will be used.\n",
    "\n",
    "* For the number blocks of the last structure, I have to choose between one, two or three **[2 Conv2D -> MaxPool2D -> Droupout]** because bigger structure takes very much time to train. I ran the code with theses structures and I have noticed that one underfits, three overfits. The better structure is: **[2 Conv2D -> MaxPool2D -> Droupout]*2**\n",
    "\n",
    "* At the end of **[2 Conv2D -> MaxPool2D -> Droupout]*2**, I choose to put a fully connected structure, I have implemented the CNN with one, two or three hidden layers. One was the more efficent."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "sxZg89oqBdun"
   },
   "source": [
    "Now, I will just check how much **Data Augmentation** makes the model more robust. So I will use the optimized CNN that I found with Grid Search on Data augmented dataset (you can check it on the code: *architecture_optimization.ipynb*)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "1NJbKwgFJtAw"
   },
   "source": [
    "# **Model with no Data Augmentation / Adam / SCE**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 881
    },
    "colab_type": "code",
    "id": "zSJ8xTO3FpQ9",
    "outputId": "971c030d-1ef8-434d-eda1-4212567087d6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 33600 samples, validate on 8400 samples\n",
      "Epoch 1/8\n",
      "33600/33600 - 510s - loss: 0.3360 - acc: 0.8932 - val_loss: 0.0778 - val_acc: 0.9777\n",
      "Epoch 2/8\n",
      "33600/33600 - 506s - loss: 0.1045 - acc: 0.9683 - val_loss: 0.0565 - val_acc: 0.9840\n",
      "Epoch 3/8\n",
      "33600/33600 - 504s - loss: 0.0787 - acc: 0.9775 - val_loss: 0.0439 - val_acc: 0.9855\n",
      "Epoch 4/8\n",
      "33600/33600 - 502s - loss: 0.0599 - acc: 0.9824 - val_loss: 0.0360 - val_acc: 0.9895\n",
      "Epoch 5/8\n",
      "33600/33600 - 507s - loss: 0.0517 - acc: 0.9852 - val_loss: 0.0382 - val_acc: 0.9889\n",
      "Epoch 6/8\n",
      "33600/33600 - 509s - loss: 0.0472 - acc: 0.9854 - val_loss: 0.0344 - val_acc: 0.9904\n",
      "Epoch 7/8\n",
      "33600/33600 - 506s - loss: 0.0419 - acc: 0.9877 - val_loss: 0.0323 - val_acc: 0.9908\n",
      "Epoch 8/8\n",
      "33600/33600 - 508s - loss: 0.0353 - acc: 0.9894 - val_loss: 0.0313 - val_acc: 0.9918\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlkAAAJCCAYAAADz6dIfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nyRIRERGpBipZIiIiItXg/wOz3WmDoNicNQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Define CNN:\n",
    "#[Conv2D->relu]*2 with n filters -> MaxPool2D (2,2) -> Dropout -> \n",
    "#[Conv2D->relu]*2 with m filters -> MaxPool2D (2,2) -> Dropout ->\n",
    "#Flatten -> Dense with p nodes -> Dropout -> Out\n",
    "\n",
    "model = Sequential()\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = (5,5),padding = 'Same', \n",
    "                 activation ='relu', input_shape = (28,28,1)))\n",
    "model.add(Conv2D(filters = 64, kernel_size = (5,5),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(MaxPool2D(pool_size=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(MaxPool2D(pool_size=(2,2), strides=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128, activation = \"relu\"))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(10, activation = \"softmax\")) #10 possible outputs because it is a 10-class problem\n",
    "\n",
    "# Compile the model\n",
    "model.compile(optimizer = 'adam' , loss = \"sparse_categorical_crossentropy\", metrics=[\"accuracy\"])\n",
    "\n",
    "epochs = 8\n",
    "batch_size = 100\n",
    "\n",
    "# Split the train and the validation set for the fitting\n",
    "X_train, X_val, Y_train, Y_val = train_test_split(train, label, test_size = 0.2, random_state=2)\n",
    "\n",
    "# Without data augmentation: accuracy of 0.98971 on Kaggle\n",
    "history_CNN_adam = model.fit(X_train, Y_train, batch_size = batch_size, epochs = epochs, \n",
    "                    validation_data = (X_val, Y_val), verbose = 2)\n",
    "\n",
    "training_loss = history_CNN_adam.history['loss']\n",
    "validation_loss = history_CNN_adam.history['val_loss']\n",
    "training_acc = history_CNN_adam.history['acc']\n",
    "validation_acc = history_CNN_adam.history['val_acc']\n",
    "\n",
    "# Plot the loss and accuracy curves for training and validation \n",
    "fig, ax = plt.subplots(2,1,figsize=(10,10))\n",
    "ax[0].plot(training_loss, color='b', label=\"Training loss\")\n",
    "ax[0].plot(validation_loss, color='r', label=\"validation loss\",axes =ax[0])\n",
    "legend = ax[0].legend(loc='best', shadow=True)\n",
    "\n",
    "ax[1].plot(training_acc, color='b', label=\"Training accuracy\")\n",
    "ax[1].plot(validation_acc, color='r',label=\"Validation accuracy\")\n",
    "legend = ax[1].legend(loc='best', shadow=True)\n",
    "\n",
    "# predict results\n",
    "sub = model.predict(test)\n",
    "\n",
    "# select the indix with the maximum probability\n",
    "sub = np.argmax(sub,axis = 1)\n",
    "\n",
    "sub_df = pd.Series(sub,name=\"Label\")\n",
    "\n",
    "submission = pd.concat([pd.Series(range(1,test.shape[0]+1), name = \"ImageId\"), sub_df],axis = 1)\n",
    "\n",
    "submission.to_csv(\"ismail_CNN_nda_adam_sce_conv_64.csv\",index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "JKqMyOcVEdT0"
   },
   "source": [
    "When I submit the result on Kaggle, I get: 0.98985 accuracy and the CV accuracy is 0.9918."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "U2ZvpfnEbUrv"
   },
   "source": [
    "# **Model with Data Augmentation / Adam / SCE**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 420
    },
    "colab_type": "code",
    "id": "6Sn1YaRLbUe-",
    "outputId": "6e1d2240-cdcc-457b-c54e-11fb8b0ec8d1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 512s 15ms/sample - loss: 0.2483 - acc: 0.9201 - val_loss: 0.0774 - val_acc: 0.9771\n",
      "Epoch 2\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 514s 15ms/sample - loss: 0.0922 - acc: 0.9736 - val_loss: 0.0526 - val_acc: 0.9843\n",
      "Epoch 3\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 514s 15ms/sample - loss: 0.0706 - acc: 0.9789 - val_loss: 0.0363 - val_acc: 0.9896\n",
      "Epoch 4\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 512s 15ms/sample - loss: 0.0581 - acc: 0.9831 - val_loss: 0.0386 - val_acc: 0.9886\n",
      "Epoch 5\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 513s 15ms/sample - loss: 0.0510 - acc: 0.9844 - val_loss: 0.0343 - val_acc: 0.9905\n",
      "Epoch 6\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 512s 15ms/sample - loss: 0.0422 - acc: 0.9876 - val_loss: 0.0293 - val_acc: 0.9908\n",
      "Epoch 7\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 515s 15ms/sample - loss: 0.0396 - acc: 0.9884 - val_loss: 0.0368 - val_acc: 0.9899\n",
      "Epoch 8\n",
      "Train on 33600 samples, validate on 8400 samples\n",
      "33600/33600 [==============================] - 522s 16ms/sample - loss: 0.0364 - acc: 0.9891 - val_loss: 0.0296 - val_acc: 0.9926\n"
     ]
    }
   ],
   "source": [
    "# Define CNN:\n",
    "#[Conv2D->relu]*2 with n filters -> MaxPool2D (2,2) -> Dropout -> \n",
    "#[Conv2D->relu]*2 with m filters -> MaxPool2D (2,2) -> Dropout ->\n",
    "#Flatten -> Dense with p nodes -> Dropout -> Out\n",
    "\n",
    "model = Sequential()\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = (5,5),padding = 'Same', \n",
    "                 activation ='relu', input_shape = (28,28,1)))\n",
    "model.add(Conv2D(filters = 64, kernel_size = (5,5),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(MaxPool2D(pool_size=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(Conv2D(filters = 64, kernel_size = (3,3),padding = 'Same', \n",
    "                 activation ='relu'))\n",
    "model.add(MaxPool2D(pool_size=(2,2), strides=(2,2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128, activation = \"relu\"))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(10, activation = \"softmax\")) #10 possible outputs because it is a 10-class problem\n",
    "\n",
    "# Compile the model\n",
    "model.compile(optimizer = 'Adam' , loss = \"sparse_categorical_crossentropy\", metrics=[\"accuracy\"])\n",
    "\n",
    "epochs = 8\n",
    "batch_size = 100\n",
    "\n",
    "# Split the train and the validation set for the fitting\n",
    "X_train, X_val, Y_train, Y_val = train_test_split(train, label, test_size = 0.2, random_state=2)\n",
    "\n",
    "# Manual data augmentation\n",
    "\n",
    "datagen = ImageDataGenerator(rotation_range=20,\n",
    "                             zoom_range=0.2\n",
    "                             )\n",
    "\n",
    "datagen.fit(X_train)\n",
    "\n",
    "training_loss = []\n",
    "validation_loss = []\n",
    "training_acc = []\n",
    "validation_acc = []\n",
    "\n",
    "# With data augmentation: accuracy of 0.9920 on Kaggle\n",
    "\n",
    "for e in range(1,epochs+1):\n",
    "\n",
    "    print('Epoch', e)\n",
    "    batches = 0\n",
    "\n",
    "    for x_batch, y_batch in datagen.flow(X_train, Y_train, batch_size=batch_size):\n",
    "\n",
    "        x_train, x_val, y_train, y_val = train_test_split(x_batch, y_batch, test_size = 0.2, random_state=2)\n",
    "        history_CNN_adam_da = model.fit(X_train, Y_train, validation_data = (X_val, Y_val))\n",
    "        training_loss.append(history_CNN_adam_da.history['loss'])\n",
    "        validation_loss.append(history_CNN_adam_da.history['val_loss'])\n",
    "        training_acc.append(history_CNN_adam_da.history['acc'])\n",
    "        validation_acc.append(history_CNN_adam_da.history['val_acc'])\n",
    "        batches += 1\n",
    "\n",
    "        if batches >= 1:\n",
    "            # we need to break the loop by hand because\n",
    "            # the generator loops indefinitely\n",
    "            break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 592
    },
    "colab_type": "code",
    "id": "c3BpBuI-SLaF",
    "outputId": "5606796e-08ab-49da-a958-0af4fd10e5bd"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlkAAAI/CAYAAABEVcwAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the loss and accuracy curves for training and validation\n",
    "\n",
    "training_loss = np.array(training_loss)\n",
    "validation_loss = np.array(validation_loss)\n",
    "training_acc = np.array(training_acc)\n",
    "validation_acc = np.array(validation_acc)\n",
    "\n",
    "fig, ax = plt.subplots(2,1,figsize=(10,10))\n",
    "\n",
    "#Plot loss curves\n",
    "\n",
    "ax[0].plot(training_loss, color='b', label=\"Training Loss\")\n",
    "ax[0].plot(validation_loss, color='r', label=\"Validation Loss\",axes =ax[0])\n",
    "legend = ax[0].legend(loc='best', shadow=True)\n",
    "\n",
    "# Plot accuracy curves\n",
    "ax[1].plot(training_acc, color='b', label=\"Training Accuracy\")\n",
    "ax[1].plot(validation_acc, color='r',label=\"Validation Accuracy\")\n",
    "legend = ax[1].legend(loc='best', shadow=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "NZMYqsBSSNrD"
   },
   "outputs": [],
   "source": [
    "# predict results\n",
    "sub = model.predict(test)\n",
    "\n",
    "# Takes the maximum probability digit\n",
    "sub = np.argmax(sub,axis = 1)\n",
    "\n",
    "#Build pandas Dataframe\n",
    "sub_df = pd.Series(sub,name=\"Label\")\n",
    "submission = pd.concat([pd.Series(range(1,test.shape[0]+1), name = \"ImageId\"), sub_df],axis = 1)\n",
    "\n",
    "#Save csv submission file to submit prediction of Kaggle\n",
    "submission.to_csv(\"ismail_CNN_da_adam_sce_optimal_conv_8.csv\",index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "6dW2Q3pIEle6"
   },
   "source": [
    "When I submit the result on Kaggle, I get: 0.99200 accuracy and the CV accuracy 0.9926. Here, I end my 2 days job. Thanks for reading !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 282
    },
    "colab_type": "code",
    "id": "aD3Vg_ndKwr1",
    "outputId": "579d3151-c008-4c84-d9f3-5fd4d3856518"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f5e90f3be10>"
      ]
     },
     "execution_count": 51,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD4CAYAAAAq5pAIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(test[0][:,:,0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 516
    },
    "colab_type": "code",
    "id": "_WK5Km5VHvDf",
    "outputId": "2ad238d1-71f1-41a6-fa86-6d376fc3fa14"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqgAAAHiCAYAAAAkkz7LAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<Figure size 864x576 with 6 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import random \n",
    "\n",
    "ax = []\n",
    "fig = plt.figure(figsize=(12, 8))\n",
    "\n",
    "for i in range(1,7):\n",
    "    random_number = random.randrange(1, 20001, 1)\n",
    "    img = test[random_number][:,:,0]\n",
    "    # create subplot and append to ax\n",
    "    ax.append( fig.add_subplot(2, 3, i) )\n",
    "    ax[-1].set_title(\"The predicted digit is: \" + str(sub[random_number]))\n",
    "    plt.subplots_adjust(hspace=0.5) \n",
    "    plt.imshow(img)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3IsZEpcxE17M"
   },
   "source": [
    "References & inspirations:\n",
    "* *Recognition of Handwritten Digit using\n",
    "Convolutional Neural Network in Python with\n",
    "Tensorflow and Comparison of Performance for\n",
    "Various Hidden Layers* by Fathma Siddique,Shadman Sakib & Md. Abu Bakr Siddique: https://arxiv.org/ftp/arxiv/papers/1909/1909.08490.pdf\n",
    "* *How to choose CNN Architecture MNIST* by Chris Deotte: https://www.kaggle.com/cdeotte/how-to-choose-cnn-architecture-mnist\n",
    "* *Introduction to CNN Keras - 0.997 (top 6%)* by Yassine Ghouzam: https://www.kaggle.com/yassineghouzam/introduction-to-cnn-keras-0-997-top-6\n"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "Code_CNN_Digits_Recognition.ipynb",
   "provenance": []
  },
  "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}