005-Transfer-Learning-Fashion.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "VTEGN4bfZheT"
   },
   "source": [
    "# Transfer Learning - Fashion MNIST\n",
    "\n",
    "Transfer Learning allows us to learn faster from a pre-trained model (on a similiar problem) by getting the feature transformation part\n",
    "\n",
    "There are two ways to do this\n",
    "- Get the prediction from the feature transformer and use that as an input to a new classifier.\n",
    "- Make the feature transformation from the pre-trained model false and only train the classifier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 34,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 2572,
     "status": "ok",
     "timestamp": 1521300228267,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "JGQqwCPRZheX",
    "outputId": "5946c5b4-c22b-4cd0-c30a-1377e3d0140d"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import keras\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "% matplotlib inline\n",
    "import vis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Get Data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import fashion_mnist\n",
    "(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = vis.fashion_mnist_label()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Step 1: Prepare the images and labels**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reshape data for convlution netowkr\n",
    "x_train_conv = x_train.reshape(x_train.shape[0], 28, 28, 1)\n",
    "x_test_conv = x_test.reshape(x_test.shape[0], 28, 28, 1)\n",
    "input_shape = (28, 28, 1)\n",
    "\n",
    "# Convert from 'uint8' to 'float32' and normalise the data to (0,1) \n",
    "x_train_conv =  x_train_conv.astype(\"float32\") / 255\n",
    "x_test_conv = x_test_conv.astype(\"float32\") / 255\n",
    "\n",
    "# convert class vectors to binary class matrices\n",
    "y_train_class = keras.utils.to_categorical(y_train, 10)\n",
    "y_test_class = keras.utils.to_categorical(y_test, 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "B8JTgp2lZhhG"
   },
   "source": [
    "Resize the image for the transfer learning model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "1BV0PfRGZhhH"
   },
   "source": [
    "Lets resize the image from (28,28,1) to (48,48,3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "a0Em5OmiZhhJ"
   },
   "outputs": [],
   "source": [
    "from skimage.transform import resize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "Ep6AxY43ZhhM"
   },
   "outputs": [],
   "source": [
    "def resize_image(image):\n",
    "    return resize(image, (96,96,3), mode=\"symmetric\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 285,
     "output_extras": [
      {
       "item_id": 1
      },
      {
       "item_id": 2
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1219,
     "status": "ok",
     "timestamp": 1521300502985,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "ZE8f-mQoZhhQ",
    "outputId": "9fcd8b02-f210-4dcc-a01c-f48b46c0afa3"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/skimage/transform/_warps.py:110: UserWarning: Anti-aliasing will be enabled by default in skimage 0.15 to avoid aliasing artifacts when down-sampling images.\n",
      "  warn(\"Anti-aliasing will be enabled by default in skimage 0.15 to \"\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(resize_image(x_train_conv[0]));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "fadjg15eZhhW"
   },
   "source": [
    "Lets resize the 10000 images in train and test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 34,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 57878,
     "status": "ok",
     "timestamp": 1521300737660,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "hFgQHU8eZhhX",
    "outputId": "19b16009-cc0a-44f4-fdfd-d828110757f6"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/skimage/transform/_warps.py:110: UserWarning: Anti-aliasing will be enabled by default in skimage 0.15 to avoid aliasing artifacts when down-sampling images.\n",
      "  warn(\"Anti-aliasing will be enabled by default in skimage 0.15 to \"\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(10000, 96, 96, 3)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%timeit\n",
    "x_train_conv_resize = np.array([resize_image(x_train_conv[x]) for x in range(x_train_conv[:10000].shape[0])])\n",
    "x_train_conv_resize.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 34,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 56175,
     "status": "ok",
     "timestamp": 1521300808676,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "WHHkM4bFZhhf",
    "outputId": "742463d3-2520-4a0b-dd0f-f2525d90f8a2"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/skimage/transform/_warps.py:110: UserWarning: Anti-aliasing will be enabled by default in skimage 0.15 to avoid aliasing artifacts when down-sampling images.\n",
      "  warn(\"Anti-aliasing will be enabled by default in skimage 0.15 to \"\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(10000, 96, 96, 3)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%timeit\n",
    "x_test_conv_resize = np.array([resize_image(x_test_conv[x]) for x in range(x_test_conv.shape[0])])\n",
    "x_test_conv_resize.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Step 2: Load a pre-trained model for the feature transfomation and add classifier model **"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3sJh1HW_Zhhm"
   },
   "source": [
    "Let us load the VGG model without the classifier layer and **do not** train the weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "F-1xVwMhZhhp"
   },
   "outputs": [],
   "source": [
    "from keras.applications.vgg16 import VGG16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "QqczTtGOZhhu"
   },
   "outputs": [],
   "source": [
    "conv_base = VGG16(include_top=False, weights='imagenet', \n",
    "                        input_tensor=None, input_shape=(96,96,3), pooling=\"avg\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "FdKE8Y8bc9g6"
   },
   "outputs": [],
   "source": [
    "conv_base.trainable = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 816,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 831,
     "status": "ok",
     "timestamp": 1521300816287,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "EZI1AcKic2Z5",
    "outputId": "60ab59fe-7c30-4b57-9384-11a6468e5e3f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 96, 96, 3)         0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 96, 96, 64)        1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 96, 96, 64)        36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 48, 48, 64)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 48, 48, 128)       73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 48, 48, 128)       147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 24, 24, 128)       0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 24, 24, 256)       295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 24, 24, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 24, 24, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 12, 12, 256)       0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 12, 12, 512)       1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 12, 12, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 12, 12, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 6, 6, 512)         0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 6, 6, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 6, 6, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 6, 6, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 3, 3, 512)         0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_1 ( (None, 512)               0         \n",
      "=================================================================\n",
      "Total params: 14,714,688\n",
      "Trainable params: 0\n",
      "Non-trainable params: 14,714,688\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "conv_base.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "size = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train_conv_resize_vgg = conv_base.predict(x_train_conv_resize[:size])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_test_conv_resize_vgg = conv_base.predict(x_test_conv_resize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_class_vgg = y_train_class[:size]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1000, 512)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train_conv_resize_vgg.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "o2AaA3fLZhh3"
   },
   "source": [
    "Lets build a Dense Layer Classifier after that to classify the fashion mnist data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "2DnTELYwZhh5"
   },
   "outputs": [],
   "source": [
    "model_transfer = Sequential()\n",
    "model_transfer.add(Dense(128, input_dim=512, activation='relu'))\n",
    "model_transfer.add(Dropout(0.25))\n",
    "model_transfer.add(Dense(64, activation='relu'))\n",
    "model_transfer.add(Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 306,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 830,
     "status": "ok",
     "timestamp": 1521300845139,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "f_ps2s5sZhh_",
    "outputId": "be3fb726-f41c-436a-8291-5b22988a4d6d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 128)               65664     \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                650       \n",
      "=================================================================\n",
      "Total params: 74,570\n",
      "Trainable params: 74,570\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model_transfer.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Step 3: Compile and fit the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "RLKx_3poZhiF"
   },
   "outputs": [],
   "source": [
    "model_transfer.compile(loss='categorical_crossentropy', optimizer=\"adam\", metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 258,
     "output_extras": [
      {
       "item_id": 184
      },
      {
       "item_id": 235
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 105236,
     "status": "ok",
     "timestamp": 1521300957552,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "fNQM59wUZhiK",
    "outputId": "cf874918-d774-4b1e-c34a-db2897dcdcdc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 0 ns, sys: 0 ns, total: 0 ns\n",
      "Wall time: 16 µs\n",
      "Train on 800 samples, validate on 200 samples\n",
      "Epoch 1/5\n",
      "800/800 [==============================] - 0s 482us/step - loss: 2.0406 - acc: 0.3150 - val_loss: 1.5352 - val_acc: 0.6000\n",
      "Epoch 2/5\n",
      "800/800 [==============================] - 0s 130us/step - loss: 1.2930 - acc: 0.6062 - val_loss: 0.9831 - val_acc: 0.6850\n",
      "Epoch 3/5\n",
      "800/800 [==============================] - 0s 130us/step - loss: 0.8955 - acc: 0.7262 - val_loss: 0.7357 - val_acc: 0.7650\n",
      "Epoch 4/5\n",
      "800/800 [==============================] - 0s 129us/step - loss: 0.7291 - acc: 0.7537 - val_loss: 0.6643 - val_acc: 0.8150\n",
      "Epoch 5/5\n",
      "800/800 [==============================] - 0s 138us/step - loss: 0.6082 - acc: 0.7887 - val_loss: 0.6405 - val_acc: 0.7450\n"
     ]
    }
   ],
   "source": [
    "%time\n",
    "output_transfer = model_transfer.fit(x_train_conv_resize_vgg, y_train_class_vgg, \n",
    "                                      batch_size=16, epochs=5, verbose=1, validation_split = 0.2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Step 4: Check the performance of the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.vegalite.v2+json": {
       "$schema": "https://vega.github.io/schema/vega-lite/v2.4.3.json",
       "config": {
        "view": {
         "height": 300,
         "width": 400
        }
       },
       "data": {
        "values": [
         {
          "batch": 1,
          "metrics": "loss",
          "type": "validation",
          "value": 1.5351784229278564
         },
         {
          "batch": 2,
          "metrics": "loss",
          "type": "validation",
          "value": 0.9831064891815186
         },
         {
          "batch": 3,
          "metrics": "loss",
          "type": "validation",
          "value": 0.7356976127624512
         },
         {
          "batch": 4,
          "metrics": "loss",
          "type": "validation",
          "value": 0.6643098258972168
         },
         {
          "batch": 5,
          "metrics": "loss",
          "type": "validation",
          "value": 0.6404746127128601
         },
         {
          "batch": 1,
          "metrics": "acc",
          "type": "validation",
          "value": 0.6
         },
         {
          "batch": 2,
          "metrics": "acc",
          "type": "validation",
          "value": 0.685
         },
         {
          "batch": 3,
          "metrics": "acc",
          "type": "validation",
          "value": 0.765
         },
         {
          "batch": 4,
          "metrics": "acc",
          "type": "validation",
          "value": 0.815
         },
         {
          "batch": 5,
          "metrics": "acc",
          "type": "validation",
          "value": 0.745
         },
         {
          "batch": 1,
          "metrics": "loss",
          "type": "training",
          "value": 2.040550079345703
         },
         {
          "batch": 2,
          "metrics": "loss",
          "type": "training",
          "value": 1.2929707050323487
         },
         {
          "batch": 3,
          "metrics": "loss",
          "type": "training",
          "value": 0.8955097824335099
         },
         {
          "batch": 4,
          "metrics": "loss",
          "type": "training",
          "value": 0.7290955901145935
         },
         {
          "batch": 5,
          "metrics": "loss",
          "type": "training",
          "value": 0.6081768244504928
         },
         {
          "batch": 1,
          "metrics": "acc",
          "type": "training",
          "value": 0.315
         },
         {
          "batch": 2,
          "metrics": "acc",
          "type": "training",
          "value": 0.60625
         },
         {
          "batch": 3,
          "metrics": "acc",
          "type": "training",
          "value": 0.72625
         },
         {
          "batch": 4,
          "metrics": "acc",
          "type": "training",
          "value": 0.75375
         },
         {
          "batch": 5,
          "metrics": "acc",
          "type": "training",
          "value": 0.78875
         }
        ]
       },
       "facet": {
        "column": {
         "field": "metrics",
         "type": "nominal"
        }
       },
       "resolve": {
        "scale": {
         "color": "shared",
         "x": "shared",
         "y": "independent"
        }
       },
       "spec": {
        "layer": [
         {
          "encoding": {
           "color": {
            "field": "type",
            "type": "nominal"
           },
           "tooltip": [
            {
             "field": "batch",
             "type": "quantitative"
            },
            {
             "field": "value",
             "type": "quantitative"
            }
           ],
           "x": {
            "field": "batch",
            "type": "quantitative"
           },
           "y": {
            "field": "value",
            "type": "quantitative"
           }
          },
          "height": 300,
          "mark": {
           "size": 50,
           "type": "circle"
          },
          "width": 300
         },
         {
          "encoding": {
           "color": {
            "field": "type",
            "type": "nominal"
           },
           "x": {
            "field": "batch",
            "type": "quantitative"
           },
           "y": {
            "field": "value",
            "type": "quantitative"
           }
          },
          "height": 300,
          "mark": "line",
          "width": 300
         }
        ]
       }
      },
      "image/png": "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",
      "text/plain": [
       "<VegaLite 2 object>\n",
       "\n",
       "If you see this message, it means the renderer has not been properly enabled\n",
       "for the frontend that you are using. For more information, see\n",
       "https://altair-viz.github.io/user_guide/troubleshooting.html\n"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vis.metrics(output_transfer.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'x_test_conv_reshape_vgg' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-25-892d47ab9747>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mscore\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel_transfer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_test_conv_reshape_vgg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_test_class\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'x_test_conv_reshape_vgg' is not defined"
     ]
    }
   ],
   "source": [
    "score = model_transfer.evaluate(x_test_conv_reshape_vgg, y_test_class, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 374,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1161,
     "status": "ok",
     "timestamp": 1521300979109,
     "user": {
      "displayName": "Amit Kapoor",
      "photoUrl": "//lh4.googleusercontent.com/-D2AEUHCgcx8/AAAAAAAAAAI/AAAAAAAAAoY/0HFe1PWxMvQ/s50-c-k-no/photo.jpg",
      "userId": "115841624057400778059"
     },
     "user_tz": -330
    },
    "id": "QtUXU4xRbgL4",
    "outputId": "0f5d659a-29f8-4e0d-dd27-9c07d558c476"
   },
   "outputs": [],
   "source": [
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Step 5: Make & Visualise the Prediction**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predict_classes_transfer = model_transfer.predict_classes(x_test_conv_reshape_vgg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "qOJk0ncjcBoa"
   },
   "outputs": [],
   "source": [
    "pd.crosstab(y_test, predict_classes_transfer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "proba_transfer = model_transfer_conv.predict_proba(x_test_conv_reshape_vgg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "vis.imshow(x_test[i], labels[y_test[i]]) | vis.predict(proba_transfer[i], y_test[i], labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "default_view": {},
   "name": "04-Transfer-Learning.ipynb",
   "provenance": [],
   "version": "0.3.2",
   "views": {}
  },
  "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
}