digit_detect_dense.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28)\n",
      "(60000,)\n",
      "[5 0 4 ... 5 6 8]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1a58bca0b08>"
      ]
     },
     "execution_count": 2,
     "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 keras\n",
    "from keras.datasets import mnist\n",
    "from keras import models, layers\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.utils import to_categorical\n",
    "\n",
    "\n",
    "# Reading the mnist dataset which is 28X28 pixel of 70k integer (0-9) RBG scale images\n",
    "# Out of these 70K, I am using 60K to train my deep network, rest 10k will be used for validation/testing\n",
    "(train_images, train_labels), (test_images, test_labels) = mnist.load_data()\n",
    "print (train_images.shape);\n",
    "print (train_labels.shape);\n",
    "print (train_labels);\n",
    "\n",
    "# Lets see the images visually using MatPlot\n",
    "plt.figure()\n",
    "plt.title('First Train Image:');\n",
    "plt.imshow(train_images[0]);\n",
    "plt.colorbar()\n",
    "plt.grid(False)\n",
    "plt.show();\n",
    "plt.title('First Test Image:');\n",
    "plt.imshow(test_images[0])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\pradipm\\AppData\\Local\\Continuum\\anaconda3\\envs\\practice\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 512)               401920    \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 407,050\n",
      "Trainable params: 407,050\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "## Building the TensorFlow Network\n",
    "# As this is a categorical classification (multi class, single label) problem, so going with seq \"Dense\" sort of deep networks\n",
    "network = models.Sequential();\n",
    "\n",
    "# First layer is downshaping 28X28 image to a 512 tensor vector, relu is max(0, input)\n",
    "network.add(layers.Dense(512, activation='relu', input_shape=(28 * 28,)));\n",
    "\n",
    "# Now adding the last layer downshaping the output of previous layer to 10 (i.e. 0-9) \"digits\".\n",
    "# Remember: the output of this layer is merely a probablistic distribution, hence \"softmax\".\n",
    "network.add(layers.Dense(10, activation='softmax'))\n",
    "\n",
    "## lets see how DL network looks like\n",
    "network.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Lets compile the network\n",
    "# Well: compile here means setting up the back propagation, define the loss function and to determine how the model is working\n",
    "## Optimizer is to reduce the loss function every step, gradient descent, reducing d(loss_fn)/d(layer's weight)\n",
    "# As it's a categocial classification, so picking up usual cat_crossent as loss method\n",
    "# How my network is converging? Lets measure 'accuracy'\n",
    "network.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy']);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Data preparation: To fit to my model\n",
    "\n",
    "train_images = train_images.reshape((60000, 28 * 28))\n",
    "train_images = train_images.astype('float32') / 255\n",
    "\n",
    "test_images = test_images.reshape((10000, 28 * 28))\n",
    "test_images = test_images.astype('float32') / 255\n",
    "\n",
    "# Sort of encoding such that the value remains b/w [-1, to +1]\n",
    "train_labels = to_categorical(train_labels)\n",
    "test_labels = to_categorical(test_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\pradipm\\AppData\\Local\\Continuum\\anaconda3\\envs\\practice\\lib\\site-packages\\tensorflow\\python\\ops\\math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 50000 samples, validate on 10000 samples\n",
      "Epoch 1/5\n",
      "50000/50000 [==============================] - 7s 145us/step - loss: 0.2822 - acc: 0.9176 - val_loss: 0.1774 - val_acc: 0.9444\n",
      "Epoch 2/5\n",
      "50000/50000 [==============================] - 6s 125us/step - loss: 0.1157 - acc: 0.9660 - val_loss: 0.1129 - val_acc: 0.9642\n",
      "Epoch 3/5\n",
      "50000/50000 [==============================] - 6s 124us/step - loss: 0.0758 - acc: 0.9775 - val_loss: 0.0860 - val_acc: 0.9750\n",
      "Epoch 4/5\n",
      "50000/50000 [==============================] - 6s 125us/step - loss: 0.0543 - acc: 0.9836 - val_loss: 0.0893 - val_acc: 0.9735\n",
      "Epoch 5/5\n",
      "50000/50000 [==============================] - 6s 124us/step - loss: 0.0399 - acc: 0.9880 - val_loss: 0.0791 - val_acc: 0.9760394 - acc\n"
     ]
    }
   ],
   "source": [
    "# Now Training is done with 5 steps \"epoch\" on training data\n",
    "# Batch size typically power of 2, it's the per batch processing in a epoch\n",
    "validation_images = train_images[:10000]\n",
    "actual_train_images = train_images[10000:]\n",
    "validation_labels = train_labels[:10000]\n",
    "actual_train_labels = train_labels[10000:]\n",
    "history = network.fit(actual_train_images, actual_train_labels, \n",
    "                      epochs=5, \n",
    "                      batch_size=128, \n",
    "                      validation_data=(validation_images, validation_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "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": [
    "## Let's show the overfitting example\n",
    "\n",
    "# history object, output of fit methof\n",
    "his_dict = history.history\n",
    "print (his_dict.keys())\n",
    "\n",
    "# loss plotting: Taining vs Validation\n",
    "loss_values = his_dict['loss']\n",
    "val_loss_values = his_dict['val_loss']\n",
    "epochs = range(1, len(his_dict['acc']) + 1)\n",
    "plt.plot(epochs, loss_values, 'b', label='Training Loss')\n",
    "plt.plot(epochs, val_loss_values, 'g', label='Validation Loss')\n",
    "plt.title('Training & Validation Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "plt.clf()   ## clear the plot\n",
    "\n",
    "# Accuracy plotting: Training vs Validation\n",
    "acc_values = his_dict['acc']\n",
    "val_acc_values = his_dict['val_acc']\n",
    "plt.plot(epochs, acc_values, 'b', label='Training Accuracy')\n",
    "plt.plot(epochs, val_acc_values, 'g', label='Validation Accuracy')\n",
    "plt.title('Training & Validation Accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 1s 71us/step\n",
      "The accuracy I am expecting: 0.9779\n"
     ]
    }
   ],
   "source": [
    "# Now it's QA time! Test it over 10k samples.\n",
    "\n",
    "test_loss, test_acc = network.evaluate(test_images, test_labels)\n",
    "print('The accuracy I am expecting:', test_acc);\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 10)\n",
      "Probablistic distribution of the first digit prediction: \n",
      "[5.2098725e-09 4.2153975e-10 9.0445951e-07 3.0629399e-05 4.7611303e-12\n",
      " 4.2860414e-08 2.3253760e-14 9.9996722e-01 1.2322017e-08 1.1891833e-06]\n",
      "Predicted pic: 7\n",
      "Actual picture under test: 7\n",
      "Bingo!\n"
     ]
    }
   ],
   "source": [
    "# Prediction section\n",
    "\n",
    "predictions = network.predict(test_images);\n",
    "print (predictions.shape);\n",
    "print ('Probablistic distribution of the first digit prediction: \\n' + str(predictions[0]));\n",
    "\n",
    "\n",
    "# See what we got?\n",
    "predicted_value = np.argmax(predictions[0]);\n",
    "print ('Predicted pic: ' + str(predicted_value));\n",
    "print ('Actual picture under test: ' + str(np.argmax(test_labels[0])));\n",
    "\n",
    "\n",
    "print ('Bingo!')"
   ]
  },
  {
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}