06b-cnns-transfer-answers.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# The Fashion MNIST - transfer learning\n",
    "This is the last time we will see the **fashion MNIST dataset**. This time we will use a pretrained CNN to process the images. For every image we want to classify it into **10 distinct clothing categories**. We will use the first layers from a known CNN, **ResNet50** and only use 10% of the dataset to speed up processing.\n",
    "\n",
    "The **ResNet50** expects the input to have **3 colors**, but the Fashion MNIST dataset is only grayscale. We will \"fix\" this issue by copying the grayscale values as the colors Red, Green, Blue. These images will undboutably look strange, but this allows us to use the ResNet50 model. Furthermore, the minimal image input size is **32 by 32** pixels, and our images are 28 by 28. To fix this, we will resize our images. Thus, our training set will have dimensions `(6000, 32, 32, 1)`. Lastly, we will do our normal **Min/Max scaling** and map the labels to their **one-hot encoding representation**. \n",
    "\n",
    "We will then download the ResNet50 model and chop off the last layers. We will then construct a new model which uses the first layers of the ResNet50 as the input layer and **freeze** the weights, so we do not train them. After feeding the input to the ResNet50 we will use the computed representation as input to a **dense layer** and then to a **softmax layer**. This will allow us to output a 10 class probability distribution specific to our problem and train the CNN using **categorical_crossentropy**. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's load the data set and inspect the dimensions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>60000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 60000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 60000\n",
       "2. 28\n",
       "3. 28\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 60000    28    28"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "60000"
      ],
      "text/latex": [
       "60000"
      ],
      "text/markdown": [
       "60000"
      ],
      "text/plain": [
       "[1] 60000"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>10000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 10000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 10000\n",
       "2. 28\n",
       "3. 28\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 10000    28    28"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "10000"
      ],
      "text/latex": [
       "10000"
      ],
      "text/markdown": [
       "10000"
      ],
      "text/plain": [
       "[1] 10000"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "library(tensorflow)\n",
    "library(keras)\n",
    "library(abind)\n",
    "source(\"06-helpers.R\")\n",
    "\n",
    "use_multi_cpu(4L)\n",
    "\n",
    "data <- dataset_fashion_mnist()\n",
    "dim(data$train$x)\n",
    "length(data$train$y)\n",
    "dim(data$test$x)\n",
    "length(data$test$y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.1 Reshape and subsampling\n",
    "For our purposes we will only work with a small number of examples this time, 6000 training examples and 1000 testing examples. Otherwise processing times might become too long. Our first step is to reshape the data, to add one extra dimension for the colors. We reshape a subsample of the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>6000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "\t<li>1</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 6000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\item 1\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 6000\n",
       "2. 28\n",
       "3. 28\n",
       "4. 1\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 6000   28   28    1"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>1000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "\t<li>1</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 1000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\item 1\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 1000\n",
       "2. 28\n",
       "3. 28\n",
       "4. 1\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 1000   28   28    1"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x_train_reshaped <- array_reshape(x = data$train$x[1:6000,,], dim = c(6000, 28, 28, 1))\n",
    "x_test_reshaped <- array_reshape(x = data$test$x[1:1000,,], dim = c(1000, 28, 28, 1))\n",
    "dim(x_train_reshaped)\n",
    "dim(x_test_reshaped)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.2 Stacking/concatenating arrays\n",
    "Since we only have one color value now and the ResNet expects 3 colors, we will copy the color value we have two more times. We do this by stacking the original color values together. To understand what is happening, think of a book, the book represents a single black and white picture. Then duplicate the book until we have 3 copies. Then we glue them together to create one book which is three times thicker by stacking the books in order. This is effectively what we are doing.\n",
    "\n",
    "To do this we use the function `abind`. Run the code below to get an idea how this works."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "5"
      ],
      "text/latex": [
       "5"
      ],
      "text/markdown": [
       "5"
      ],
      "text/plain": [
       "[1] 5"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>1</li>\n",
       "\t<li>2</li>\n",
       "\t<li>3</li>\n",
       "\t<li>4</li>\n",
       "\t<li>5</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 1\n",
       "\\item 2\n",
       "\\item 3\n",
       "\\item 4\n",
       "\\item 5\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 1\n",
       "2. 2\n",
       "3. 3\n",
       "4. 4\n",
       "5. 5\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 1 2 3 4 5"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "10"
      ],
      "text/latex": [
       "10"
      ],
      "text/markdown": [
       "10"
      ],
      "text/plain": [
       "[1] 10"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>1</li>\n",
       "\t<li>2</li>\n",
       "\t<li>3</li>\n",
       "\t<li>4</li>\n",
       "\t<li>5</li>\n",
       "\t<li>1</li>\n",
       "\t<li>2</li>\n",
       "\t<li>3</li>\n",
       "\t<li>4</li>\n",
       "\t<li>5</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 1\n",
       "\\item 2\n",
       "\\item 3\n",
       "\\item 4\n",
       "\\item 5\n",
       "\\item 1\n",
       "\\item 2\n",
       "\\item 3\n",
       "\\item 4\n",
       "\\item 5\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 1\n",
       "2. 2\n",
       "3. 3\n",
       "4. 4\n",
       "5. 5\n",
       "6. 1\n",
       "7. 2\n",
       "8. 3\n",
       "9. 4\n",
       "10. 5\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       " [1] 1 2 3 4 5 1 2 3 4 5"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "vector <- array(1:5, dim = c(5))\n",
    "dim(vector)\n",
    "vector\n",
    "long_vector <- abind(vector, vector)\n",
    "dim(long_vector)\n",
    "long_vector"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You can see that we just stacked/concatinated this vector to create a longer vector. Let's now try stacking these vectors so that we add the other vector as a column."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>5</li>\n",
       "\t<li>1</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 5\n",
       "\\item 1\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 5\n",
       "2. 1\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 5 1"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<tbody>\n",
       "\t<tr><td>1</td></tr>\n",
       "\t<tr><td>2</td></tr>\n",
       "\t<tr><td>3</td></tr>\n",
       "\t<tr><td>4</td></tr>\n",
       "\t<tr><td>5</td></tr>\n",
       "</tbody>\n",
       "</table>\n"
      ],
      "text/latex": [
       "\\begin{tabular}{l}\n",
       "\t 1\\\\\n",
       "\t 2\\\\\n",
       "\t 3\\\\\n",
       "\t 4\\\\\n",
       "\t 5\\\\\n",
       "\\end{tabular}\n"
      ],
      "text/markdown": [
       "\n",
       "| 1 |\n",
       "| 2 |\n",
       "| 3 |\n",
       "| 4 |\n",
       "| 5 |\n",
       "\n"
      ],
      "text/plain": [
       "     [,1]\n",
       "[1,] 1   \n",
       "[2,] 2   \n",
       "[3,] 3   \n",
       "[4,] 4   \n",
       "[5,] 5   "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>5</li>\n",
       "\t<li>2</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 5\n",
       "\\item 2\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 5\n",
       "2. 2\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 5 2"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<tbody>\n",
       "\t<tr><td>1</td><td>1</td></tr>\n",
       "\t<tr><td>2</td><td>2</td></tr>\n",
       "\t<tr><td>3</td><td>3</td></tr>\n",
       "\t<tr><td>4</td><td>4</td></tr>\n",
       "\t<tr><td>5</td><td>5</td></tr>\n",
       "</tbody>\n",
       "</table>\n"
      ],
      "text/latex": [
       "\\begin{tabular}{ll}\n",
       "\t 1 & 1\\\\\n",
       "\t 2 & 2\\\\\n",
       "\t 3 & 3\\\\\n",
       "\t 4 & 4\\\\\n",
       "\t 5 & 5\\\\\n",
       "\\end{tabular}\n"
      ],
      "text/markdown": [
       "\n",
       "| 1 | 1 |\n",
       "| 2 | 2 |\n",
       "| 3 | 3 |\n",
       "| 4 | 4 |\n",
       "| 5 | 5 |\n",
       "\n"
      ],
      "text/plain": [
       "     [,1] [,2]\n",
       "[1,] 1    1   \n",
       "[2,] 2    2   \n",
       "[3,] 3    3   \n",
       "[4,] 4    4   \n",
       "[5,] 5    5   "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "two_dimension <- array(1:5, dim = c(5, 1))\n",
    "dim(two_dimension)\n",
    "two_dimension\n",
    "two_dimension_duplicated <- abind(two_dimension, two_dimension)\n",
    "dim(two_dimension_duplicated)\n",
    "two_dimension_duplicated"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we added the vector as a new column. This then generalises to more dimensions, and in the cell below we just stack our reshaped data along the last axis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>6000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "\t<li>1</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 6000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\item 1\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 6000\n",
       "2. 28\n",
       "3. 28\n",
       "4. 1\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 6000   28   28    1"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>6000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "\t<li>3</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 6000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\item 3\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 6000\n",
       "2. 28\n",
       "3. 28\n",
       "4. 3\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 6000   28   28    3"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>1000</li>\n",
       "\t<li>28</li>\n",
       "\t<li>28</li>\n",
       "\t<li>3</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 1000\n",
       "\\item 28\n",
       "\\item 28\n",
       "\\item 3\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 1000\n",
       "2. 28\n",
       "3. 28\n",
       "4. 3\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 1000   28   28    3"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dim(x_train_reshaped)\n",
    "x_train_stacked <- abind(x_train_reshaped, x_train_reshaped, x_train_reshaped)\n",
    "x_test_stacked <- abind(x_test_reshaped, x_test_reshaped, x_test_reshaped)\n",
    "dim(x_train_stacked)\n",
    "dim(x_test_stacked)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.3 Resizing\n",
    "Now we need to make our images slightly larger by resizing them to 32 by 32 pixels. This is the minimal image size accepted. We use the `image_array_resize` function provided with Keras for this.\n",
    "\n",
    "This cell will take about 50 seconds to run."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Time difference of 38.97357 secs"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>6000</li>\n",
       "\t<li>32</li>\n",
       "\t<li>32</li>\n",
       "\t<li>3</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 6000\n",
       "\\item 32\n",
       "\\item 32\n",
       "\\item 3\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 6000\n",
       "2. 32\n",
       "3. 32\n",
       "4. 3\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 6000   32   32    3"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>1000</li>\n",
       "\t<li>32</li>\n",
       "\t<li>32</li>\n",
       "\t<li>3</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 1000\n",
       "\\item 32\n",
       "\\item 32\n",
       "\\item 3\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 1000\n",
       "2. 32\n",
       "3. 32\n",
       "4. 3\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 1000   32   32    3"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Just to time the function\n",
    "start <- Sys.time()\n",
    "\n",
    "# Initialise our output arrays with the correct dimensions\n",
    "x_train_resized <- array(0, dim = c(6000, 32, 32, 3))\n",
    "x_test_resized <- array(0, dim = c(1000, 32, 32, 3))\n",
    "\n",
    "# We iterate through the image sequentially - poor performance\n",
    "for (image_index in 1:dim(x_train_stacked)[1]) {\n",
    "    x_train_resized[image_index,,,] <- image_array_resize(x_train_stacked[image_index,,,],\n",
    "                                                          32, 32, data_format = c(\"channels_last\"))\n",
    "}\n",
    "for (image_index in 1:dim(x_test_stacked)[1]) {\n",
    "    x_test_resized[image_index,,,] <- image_array_resize(x_test_stacked[image_index,,,],\n",
    "                                                          32, 32, data_format = c(\"channels_last\"))\n",
    "}\n",
    "\n",
    "end <- Sys.time()\n",
    "end - start\n",
    "# Check our work\n",
    "dim(x_train_resized)\n",
    "dim(x_test_resized)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.4 Scaling\n",
    "Next we scale the image using the Min/Max approach like before. Since all our features are pixel values, we just divide each feature by 255."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train <- x_train_resized/255\n",
    "x_test <- x_test_resized/255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.5 One-hot encoding\n",
    "Lastly, the one-hot encoding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>6000</li>\n",
       "\t<li>10</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 6000\n",
       "\\item 10\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 6000\n",
       "2. 10\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 6000   10"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<ol class=list-inline>\n",
       "\t<li>1000</li>\n",
       "\t<li>10</li>\n",
       "</ol>\n"
      ],
      "text/latex": [
       "\\begin{enumerate*}\n",
       "\\item 1000\n",
       "\\item 10\n",
       "\\end{enumerate*}\n"
      ],
      "text/markdown": [
       "1. 1000\n",
       "2. 10\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "[1] 1000   10"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_train = to_categorical(y = data$train$y[1:6000], num_classes = 10)\n",
    "y_test = to_categorical(y = data$test$y[1:1000], num_classes = 10)\n",
    "dim(y_train)\n",
    "dim(y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.1 ResNet50\n",
    "Now we want to use the first layers of the ResNet50 model, pretrained, as feature extractors for our final model. We start by downloading the model description along with the weights. \n",
    "\n",
    "The ResNet50 model which we download has been trained with a known dataset called [ImageNet](http://www.image-net.org/). This is a very large dataset (14 million images) thus we can benefitted from someone else spending a long time training on this dataset. The model was trained in the multiclass classification task over 1000 classes. We do not care to classify our clothing images as these classes so we specify `include_top = FALSE`, to indicate that we will not use the last layer of the model, which is the softmax layer over 1000 classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Time difference of 6.701274 secs"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "start <- Sys.time()\n",
    "\n",
    "k_clear_session()\n",
    "resnet_50_model <- application_resnet50(weights = 'imagenet', include_top = FALSE, input_shape = c(32, 32, 3))\n",
    "\n",
    "end <- Sys.time()\n",
    "end - start"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us check the layers on this model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "Layer (type)              Output Shape      Param #  Connected to               \n",
      "================================================================================\n",
      "input_1 (InputLayer)      (None, 32, 32, 3) 0                                   \n",
      "________________________________________________________________________________\n",
      "conv1_pad (ZeroPadding2D) (None, 38, 38, 3) 0        input_1[0][0]              \n",
      "________________________________________________________________________________\n",
      "conv1 (Conv2D)            (None, 16, 16, 64 9472     conv1_pad[0][0]            \n",
      "________________________________________________________________________________\n",
      "bn_conv1 (BatchNormalizat (None, 16, 16, 64 256      conv1[0][0]                \n",
      "________________________________________________________________________________\n",
      "activation (Activation)   (None, 16, 16, 64 0        bn_conv1[0][0]             \n",
      "________________________________________________________________________________\n",
      "pool1_pad (ZeroPadding2D) (None, 18, 18, 64 0        activation[0][0]           \n",
      "________________________________________________________________________________\n",
      "max_pooling2d (MaxPooling (None, 8, 8, 64)  0        pool1_pad[0][0]            \n",
      "________________________________________________________________________________\n",
      "res2a_branch2a (Conv2D)   (None, 8, 8, 64)  4160     max_pooling2d[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn2a_branch2a (BatchNorma (None, 8, 8, 64)  256      res2a_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_1 (Activation) (None, 8, 8, 64)  0        bn2a_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2a_branch2b (Conv2D)   (None, 8, 8, 64)  36928    activation_1[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2a_branch2b (BatchNorma (None, 8, 8, 64)  256      res2a_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_2 (Activation) (None, 8, 8, 64)  0        bn2a_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2a_branch2c (Conv2D)   (None, 8, 8, 256) 16640    activation_2[0][0]         \n",
      "________________________________________________________________________________\n",
      "res2a_branch1 (Conv2D)    (None, 8, 8, 256) 16640    max_pooling2d[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn2a_branch2c (BatchNorma (None, 8, 8, 256) 1024     res2a_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "bn2a_branch1 (BatchNormal (None, 8, 8, 256) 1024     res2a_branch1[0][0]        \n",
      "________________________________________________________________________________\n",
      "add (Add)                 (None, 8, 8, 256) 0        bn2a_branch2c[0][0]        \n",
      "                                                     bn2a_branch1[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_3 (Activation) (None, 8, 8, 256) 0        add[0][0]                  \n",
      "________________________________________________________________________________\n",
      "res2b_branch2a (Conv2D)   (None, 8, 8, 64)  16448    activation_3[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2b_branch2a (BatchNorma (None, 8, 8, 64)  256      res2b_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_4 (Activation) (None, 8, 8, 64)  0        bn2b_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2b_branch2b (Conv2D)   (None, 8, 8, 64)  36928    activation_4[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2b_branch2b (BatchNorma (None, 8, 8, 64)  256      res2b_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_5 (Activation) (None, 8, 8, 64)  0        bn2b_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2b_branch2c (Conv2D)   (None, 8, 8, 256) 16640    activation_5[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2b_branch2c (BatchNorma (None, 8, 8, 256) 1024     res2b_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_1 (Add)               (None, 8, 8, 256) 0        bn2b_branch2c[0][0]        \n",
      "                                                     activation_3[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_6 (Activation) (None, 8, 8, 256) 0        add_1[0][0]                \n",
      "________________________________________________________________________________\n",
      "res2c_branch2a (Conv2D)   (None, 8, 8, 64)  16448    activation_6[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2c_branch2a (BatchNorma (None, 8, 8, 64)  256      res2c_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_7 (Activation) (None, 8, 8, 64)  0        bn2c_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2c_branch2b (Conv2D)   (None, 8, 8, 64)  36928    activation_7[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2c_branch2b (BatchNorma (None, 8, 8, 64)  256      res2c_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_8 (Activation) (None, 8, 8, 64)  0        bn2c_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2c_branch2c (Conv2D)   (None, 8, 8, 256) 16640    activation_8[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2c_branch2c (BatchNorma (None, 8, 8, 256) 1024     res2c_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_2 (Add)               (None, 8, 8, 256) 0        bn2c_branch2c[0][0]        \n",
      "                                                     activation_6[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_9 (Activation) (None, 8, 8, 256) 0        add_2[0][0]                \n",
      "________________________________________________________________________________\n",
      "res3a_branch2a (Conv2D)   (None, 4, 4, 128) 32896    activation_9[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn3a_branch2a (BatchNorma (None, 4, 4, 128) 512      res3a_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_10 (Activation (None, 4, 4, 128) 0        bn3a_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3a_branch2b (Conv2D)   (None, 4, 4, 128) 147584   activation_10[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3a_branch2b (BatchNorma (None, 4, 4, 128) 512      res3a_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_11 (Activation (None, 4, 4, 128) 0        bn3a_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3a_branch2c (Conv2D)   (None, 4, 4, 512) 66048    activation_11[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3a_branch1 (Conv2D)    (None, 4, 4, 512) 131584   activation_9[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn3a_branch2c (BatchNorma (None, 4, 4, 512) 2048     res3a_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "bn3a_branch1 (BatchNormal (None, 4, 4, 512) 2048     res3a_branch1[0][0]        \n",
      "________________________________________________________________________________\n",
      "add_3 (Add)               (None, 4, 4, 512) 0        bn3a_branch2c[0][0]        \n",
      "                                                     bn3a_branch1[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_12 (Activation (None, 4, 4, 512) 0        add_3[0][0]                \n",
      "________________________________________________________________________________\n",
      "res3b_branch2a (Conv2D)   (None, 4, 4, 128) 65664    activation_12[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3b_branch2a (BatchNorma (None, 4, 4, 128) 512      res3b_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_13 (Activation (None, 4, 4, 128) 0        bn3b_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3b_branch2b (Conv2D)   (None, 4, 4, 128) 147584   activation_13[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3b_branch2b (BatchNorma (None, 4, 4, 128) 512      res3b_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_14 (Activation (None, 4, 4, 128) 0        bn3b_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3b_branch2c (Conv2D)   (None, 4, 4, 512) 66048    activation_14[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3b_branch2c (BatchNorma (None, 4, 4, 512) 2048     res3b_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_4 (Add)               (None, 4, 4, 512) 0        bn3b_branch2c[0][0]        \n",
      "                                                     activation_12[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_15 (Activation (None, 4, 4, 512) 0        add_4[0][0]                \n",
      "________________________________________________________________________________\n",
      "res3c_branch2a (Conv2D)   (None, 4, 4, 128) 65664    activation_15[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3c_branch2a (BatchNorma (None, 4, 4, 128) 512      res3c_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_16 (Activation (None, 4, 4, 128) 0        bn3c_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3c_branch2b (Conv2D)   (None, 4, 4, 128) 147584   activation_16[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3c_branch2b (BatchNorma (None, 4, 4, 128) 512      res3c_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_17 (Activation (None, 4, 4, 128) 0        bn3c_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3c_branch2c (Conv2D)   (None, 4, 4, 512) 66048    activation_17[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3c_branch2c (BatchNorma (None, 4, 4, 512) 2048     res3c_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_5 (Add)               (None, 4, 4, 512) 0        bn3c_branch2c[0][0]        \n",
      "                                                     activation_15[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_18 (Activation (None, 4, 4, 512) 0        add_5[0][0]                \n",
      "________________________________________________________________________________\n",
      "res3d_branch2a (Conv2D)   (None, 4, 4, 128) 65664    activation_18[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3d_branch2a (BatchNorma (None, 4, 4, 128) 512      res3d_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_19 (Activation (None, 4, 4, 128) 0        bn3d_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3d_branch2b (Conv2D)   (None, 4, 4, 128) 147584   activation_19[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3d_branch2b (BatchNorma (None, 4, 4, 128) 512      res3d_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_20 (Activation (None, 4, 4, 128) 0        bn3d_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res3d_branch2c (Conv2D)   (None, 4, 4, 512) 66048    activation_20[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn3d_branch2c (BatchNorma (None, 4, 4, 512) 2048     res3d_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_6 (Add)               (None, 4, 4, 512) 0        bn3d_branch2c[0][0]        \n",
      "                                                     activation_18[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_21 (Activation (None, 4, 4, 512) 0        add_6[0][0]                \n",
      "________________________________________________________________________________\n",
      "res4a_branch2a (Conv2D)   (None, 2, 2, 256) 131328   activation_21[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4a_branch2a (BatchNorma (None, 2, 2, 256) 1024     res4a_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_22 (Activation (None, 2, 2, 256) 0        bn4a_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4a_branch2b (Conv2D)   (None, 2, 2, 256) 590080   activation_22[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4a_branch2b (BatchNorma (None, 2, 2, 256) 1024     res4a_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_23 (Activation (None, 2, 2, 256) 0        bn4a_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4a_branch2c (Conv2D)   (None, 2, 2, 1024 263168   activation_23[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4a_branch1 (Conv2D)    (None, 2, 2, 1024 525312   activation_21[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4a_branch2c (BatchNorma (None, 2, 2, 1024 4096     res4a_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "bn4a_branch1 (BatchNormal (None, 2, 2, 1024 4096     res4a_branch1[0][0]        \n",
      "________________________________________________________________________________\n",
      "add_7 (Add)               (None, 2, 2, 1024 0        bn4a_branch2c[0][0]        \n",
      "                                                     bn4a_branch1[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_24 (Activation (None, 2, 2, 1024 0        add_7[0][0]                \n",
      "________________________________________________________________________________\n",
      "res4b_branch2a (Conv2D)   (None, 2, 2, 256) 262400   activation_24[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4b_branch2a (BatchNorma (None, 2, 2, 256) 1024     res4b_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_25 (Activation (None, 2, 2, 256) 0        bn4b_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4b_branch2b (Conv2D)   (None, 2, 2, 256) 590080   activation_25[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4b_branch2b (BatchNorma (None, 2, 2, 256) 1024     res4b_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_26 (Activation (None, 2, 2, 256) 0        bn4b_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4b_branch2c (Conv2D)   (None, 2, 2, 1024 263168   activation_26[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4b_branch2c (BatchNorma (None, 2, 2, 1024 4096     res4b_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_8 (Add)               (None, 2, 2, 1024 0        bn4b_branch2c[0][0]        \n",
      "                                                     activation_24[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_27 (Activation (None, 2, 2, 1024 0        add_8[0][0]                \n",
      "________________________________________________________________________________\n",
      "res4c_branch2a (Conv2D)   (None, 2, 2, 256) 262400   activation_27[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4c_branch2a (BatchNorma (None, 2, 2, 256) 1024     res4c_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_28 (Activation (None, 2, 2, 256) 0        bn4c_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4c_branch2b (Conv2D)   (None, 2, 2, 256) 590080   activation_28[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4c_branch2b (BatchNorma (None, 2, 2, 256) 1024     res4c_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_29 (Activation (None, 2, 2, 256) 0        bn4c_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4c_branch2c (Conv2D)   (None, 2, 2, 1024 263168   activation_29[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4c_branch2c (BatchNorma (None, 2, 2, 1024 4096     res4c_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_9 (Add)               (None, 2, 2, 1024 0        bn4c_branch2c[0][0]        \n",
      "                                                     activation_27[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_30 (Activation (None, 2, 2, 1024 0        add_9[0][0]                \n",
      "________________________________________________________________________________\n",
      "res4d_branch2a (Conv2D)   (None, 2, 2, 256) 262400   activation_30[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4d_branch2a (BatchNorma (None, 2, 2, 256) 1024     res4d_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_31 (Activation (None, 2, 2, 256) 0        bn4d_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4d_branch2b (Conv2D)   (None, 2, 2, 256) 590080   activation_31[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4d_branch2b (BatchNorma (None, 2, 2, 256) 1024     res4d_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_32 (Activation (None, 2, 2, 256) 0        bn4d_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4d_branch2c (Conv2D)   (None, 2, 2, 1024 263168   activation_32[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4d_branch2c (BatchNorma (None, 2, 2, 1024 4096     res4d_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_10 (Add)              (None, 2, 2, 1024 0        bn4d_branch2c[0][0]        \n",
      "                                                     activation_30[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_33 (Activation (None, 2, 2, 1024 0        add_10[0][0]               \n",
      "________________________________________________________________________________\n",
      "res4e_branch2a (Conv2D)   (None, 2, 2, 256) 262400   activation_33[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4e_branch2a (BatchNorma (None, 2, 2, 256) 1024     res4e_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_34 (Activation (None, 2, 2, 256) 0        bn4e_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4e_branch2b (Conv2D)   (None, 2, 2, 256) 590080   activation_34[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4e_branch2b (BatchNorma (None, 2, 2, 256) 1024     res4e_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_35 (Activation (None, 2, 2, 256) 0        bn4e_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4e_branch2c (Conv2D)   (None, 2, 2, 1024 263168   activation_35[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4e_branch2c (BatchNorma (None, 2, 2, 1024 4096     res4e_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_11 (Add)              (None, 2, 2, 1024 0        bn4e_branch2c[0][0]        \n",
      "                                                     activation_33[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_36 (Activation (None, 2, 2, 1024 0        add_11[0][0]               \n",
      "________________________________________________________________________________\n",
      "res4f_branch2a (Conv2D)   (None, 2, 2, 256) 262400   activation_36[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4f_branch2a (BatchNorma (None, 2, 2, 256) 1024     res4f_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_37 (Activation (None, 2, 2, 256) 0        bn4f_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4f_branch2b (Conv2D)   (None, 2, 2, 256) 590080   activation_37[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4f_branch2b (BatchNorma (None, 2, 2, 256) 1024     res4f_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_38 (Activation (None, 2, 2, 256) 0        bn4f_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res4f_branch2c (Conv2D)   (None, 2, 2, 1024 263168   activation_38[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn4f_branch2c (BatchNorma (None, 2, 2, 1024 4096     res4f_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_12 (Add)              (None, 2, 2, 1024 0        bn4f_branch2c[0][0]        \n",
      "                                                     activation_36[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_39 (Activation (None, 2, 2, 1024 0        add_12[0][0]               \n",
      "________________________________________________________________________________\n",
      "res5a_branch2a (Conv2D)   (None, 1, 1, 512) 524800   activation_39[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5a_branch2a (BatchNorma (None, 1, 1, 512) 2048     res5a_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_40 (Activation (None, 1, 1, 512) 0        bn5a_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5a_branch2b (Conv2D)   (None, 1, 1, 512) 2359808  activation_40[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5a_branch2b (BatchNorma (None, 1, 1, 512) 2048     res5a_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_41 (Activation (None, 1, 1, 512) 0        bn5a_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5a_branch2c (Conv2D)   (None, 1, 1, 2048 1050624  activation_41[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5a_branch1 (Conv2D)    (None, 1, 1, 2048 2099200  activation_39[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5a_branch2c (BatchNorma (None, 1, 1, 2048 8192     res5a_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "bn5a_branch1 (BatchNormal (None, 1, 1, 2048 8192     res5a_branch1[0][0]        \n",
      "________________________________________________________________________________\n",
      "add_13 (Add)              (None, 1, 1, 2048 0        bn5a_branch2c[0][0]        \n",
      "                                                     bn5a_branch1[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_42 (Activation (None, 1, 1, 2048 0        add_13[0][0]               \n",
      "________________________________________________________________________________\n",
      "res5b_branch2a (Conv2D)   (None, 1, 1, 512) 1049088  activation_42[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5b_branch2a (BatchNorma (None, 1, 1, 512) 2048     res5b_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_43 (Activation (None, 1, 1, 512) 0        bn5b_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5b_branch2b (Conv2D)   (None, 1, 1, 512) 2359808  activation_43[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5b_branch2b (BatchNorma (None, 1, 1, 512) 2048     res5b_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_44 (Activation (None, 1, 1, 512) 0        bn5b_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5b_branch2c (Conv2D)   (None, 1, 1, 2048 1050624  activation_44[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5b_branch2c (BatchNorma (None, 1, 1, 2048 8192     res5b_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_14 (Add)              (None, 1, 1, 2048 0        bn5b_branch2c[0][0]        \n",
      "                                                     activation_42[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_45 (Activation (None, 1, 1, 2048 0        add_14[0][0]               \n",
      "________________________________________________________________________________\n",
      "res5c_branch2a (Conv2D)   (None, 1, 1, 512) 1049088  activation_45[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5c_branch2a (BatchNorma (None, 1, 1, 512) 2048     res5c_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_46 (Activation (None, 1, 1, 512) 0        bn5c_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5c_branch2b (Conv2D)   (None, 1, 1, 512) 2359808  activation_46[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5c_branch2b (BatchNorma (None, 1, 1, 512) 2048     res5c_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_47 (Activation (None, 1, 1, 512) 0        bn5c_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res5c_branch2c (Conv2D)   (None, 1, 1, 2048 1050624  activation_47[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn5c_branch2c (BatchNorma (None, 1, 1, 2048 8192     res5c_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_15 (Add)              (None, 1, 1, 2048 0        bn5c_branch2c[0][0]        \n",
      "                                                     activation_45[0][0]        \n",
      "________________________________________________________________________________\n",
      "activation_48 (Activation (None, 1, 1, 2048 0        add_15[0][0]               \n",
      "================================================================================\n",
      "Total params: 23,587,712\n",
      "Trainable params: 23,534,592\n",
      "Non-trainable params: 53,120\n",
      "________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "summary(resnet_50_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Wow! This is a very big model, 23 million parameters with lots of layers. In fact, processing our images through this network takes some time and we also expect the first features learned by the network to be beneficial to our problem so in the next cell we only use the layers up to the layer `activation_6`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "smaller_resnet_50 <- keras_model(inputs = resnet_50_model$input, \n",
    "                                 outputs = get_layer(resnet_50_model, 'activation_6')$output)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Furthermore, we do not want to train those layers. Let's make sure that we do not train all of these layers by freezing the weights."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "Layer (type)              Output Shape      Param #  Connected to               \n",
      "================================================================================\n",
      "input_1 (InputLayer)      (None, 32, 32, 3) 0                                   \n",
      "________________________________________________________________________________\n",
      "conv1_pad (ZeroPadding2D) (None, 38, 38, 3) 0        input_1[0][0]              \n",
      "________________________________________________________________________________\n",
      "conv1 (Conv2D)            (None, 16, 16, 64 9472     conv1_pad[0][0]            \n",
      "________________________________________________________________________________\n",
      "bn_conv1 (BatchNormalizat (None, 16, 16, 64 256      conv1[0][0]                \n",
      "________________________________________________________________________________\n",
      "activation (Activation)   (None, 16, 16, 64 0        bn_conv1[0][0]             \n",
      "________________________________________________________________________________\n",
      "pool1_pad (ZeroPadding2D) (None, 18, 18, 64 0        activation[0][0]           \n",
      "________________________________________________________________________________\n",
      "max_pooling2d (MaxPooling (None, 8, 8, 64)  0        pool1_pad[0][0]            \n",
      "________________________________________________________________________________\n",
      "res2a_branch2a (Conv2D)   (None, 8, 8, 64)  4160     max_pooling2d[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn2a_branch2a (BatchNorma (None, 8, 8, 64)  256      res2a_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_1 (Activation) (None, 8, 8, 64)  0        bn2a_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2a_branch2b (Conv2D)   (None, 8, 8, 64)  36928    activation_1[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2a_branch2b (BatchNorma (None, 8, 8, 64)  256      res2a_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_2 (Activation) (None, 8, 8, 64)  0        bn2a_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2a_branch2c (Conv2D)   (None, 8, 8, 256) 16640    activation_2[0][0]         \n",
      "________________________________________________________________________________\n",
      "res2a_branch1 (Conv2D)    (None, 8, 8, 256) 16640    max_pooling2d[0][0]        \n",
      "________________________________________________________________________________\n",
      "bn2a_branch2c (BatchNorma (None, 8, 8, 256) 1024     res2a_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "bn2a_branch1 (BatchNormal (None, 8, 8, 256) 1024     res2a_branch1[0][0]        \n",
      "________________________________________________________________________________\n",
      "add (Add)                 (None, 8, 8, 256) 0        bn2a_branch2c[0][0]        \n",
      "                                                     bn2a_branch1[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_3 (Activation) (None, 8, 8, 256) 0        add[0][0]                  \n",
      "________________________________________________________________________________\n",
      "res2b_branch2a (Conv2D)   (None, 8, 8, 64)  16448    activation_3[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2b_branch2a (BatchNorma (None, 8, 8, 64)  256      res2b_branch2a[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_4 (Activation) (None, 8, 8, 64)  0        bn2b_branch2a[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2b_branch2b (Conv2D)   (None, 8, 8, 64)  36928    activation_4[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2b_branch2b (BatchNorma (None, 8, 8, 64)  256      res2b_branch2b[0][0]       \n",
      "________________________________________________________________________________\n",
      "activation_5 (Activation) (None, 8, 8, 64)  0        bn2b_branch2b[0][0]        \n",
      "________________________________________________________________________________\n",
      "res2b_branch2c (Conv2D)   (None, 8, 8, 256) 16640    activation_5[0][0]         \n",
      "________________________________________________________________________________\n",
      "bn2b_branch2c (BatchNorma (None, 8, 8, 256) 1024     res2b_branch2c[0][0]       \n",
      "________________________________________________________________________________\n",
      "add_1 (Add)               (None, 8, 8, 256) 0        bn2b_branch2c[0][0]        \n",
      "                                                     activation_3[0][0]         \n",
      "________________________________________________________________________________\n",
      "activation_6 (Activation) (None, 8, 8, 256) 0        add_1[0][0]                \n",
      "================================================================================\n",
      "Total params: 158,208\n",
      "Trainable params: 0\n",
      "Non-trainable params: 158,208\n",
      "________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "freeze_weights(smaller_resnet_50)\n",
    "summary(smaller_resnet_50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can see that there are no trainable parameters in this model. Good!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.2 Our model\n",
    "Let us add a dense layer and our own softmax layer to the `smaller_resnet_50` model and start training it.\n",
    "\n",
    "Take notice of\n",
    "- The first \"layer\" in our network is the `smaller_resnet_50` model. It takes the initial input and outputs some features.\n",
    "- The small learning rate. A small learning rate is very common when training with pretrained networks.\n",
    "- Long training time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "Layer (type)                        Output Shape                    Param #     \n",
      "================================================================================\n",
      "model (Model)                       (None, 8, 8, 256)               158208      \n",
      "________________________________________________________________________________\n",
      "flatten_2 (Flatten)                 (None, 16384)                   0           \n",
      "________________________________________________________________________________\n",
      "dense_4 (Dense)                     (None, 32)                      524320      \n",
      "________________________________________________________________________________\n",
      "dense_5 (Dense)                     (None, 10)                      330         \n",
      "================================================================================\n",
      "Total params: 682,858\n",
      "Trainable params: 524,650\n",
      "Non-trainable params: 158,208\n",
      "________________________________________________________________________________\n",
      "Epoch 30 - loss: 0.1076186, acc.: 0.9720833, val loss: 0.338638777434826, val acc.: 0.889999985694885  \r"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0gAAANICAIAAAByhViMAAAACXBIWXMAABJ0AAASdAHeZh94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\n2AEAAEiCYgcAACCJ/wM8S5Ov6C6jSwAAAABJRU5ErkJggg==",
      "text/plain": [
       "plot without title"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model <- keras_model_sequential() %>%\n",
    "    smaller_resnet_50 %>%\n",
    "    layer_flatten() %>%\n",
    "    layer_dense(units = 32, activation = \"relu\") %>%\n",
    "    layer_dense(units = 10, activation = \"softmax\")\n",
    "\n",
    "cat(summary(model))\n",
    "\n",
    "model %>% compile(\n",
    "    optimizer = optimizer_adam(lr = 0.0001),\n",
    "    loss = \"categorical_crossentropy\",\n",
    "    metric = \"acc\"\n",
    ")\n",
    "\n",
    "history <- model %>% fit(\n",
    "    x = x_train,\n",
    "    y = y_train,\n",
    "    validation_split = 0.2,\n",
    "    epochs = 20,\n",
    "    batch_size = 64,\n",
    "    callbacks=list(Progress$new())\n",
    ")\n",
    "\n",
    "plot(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the plot we can see that our model is fairly expressive and we could maybe continue to train it and seen some performance improvements but we leave this is a bonus exercise."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<dl>\n",
       "\t<dt>$loss</dt>\n",
       "\t\t<dd>3.55828490447998</dd>\n",
       "\t<dt>$acc</dt>\n",
       "\t\t<dd>0.0949999988079071</dd>\n",
       "</dl>\n"
      ],
      "text/latex": [
       "\\begin{description}\n",
       "\\item[\\$loss] 3.55828490447998\n",
       "\\item[\\$acc] 0.0949999988079071\n",
       "\\end{description}\n"
      ],
      "text/markdown": [
       "$loss\n",
       ":   3.55828490447998\n",
       "$acc\n",
       ":   0.0949999988079071\n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "$loss\n",
       "[1] 3.558285\n",
       "\n",
       "$acc\n",
       "[1] 0.095\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model %>% evaluate(x_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this notebook we saw how to use a pretrained CNN and how we could add our layers to the model so that we can benefit from the generic features learned by the model. Some notes about this notebook.\n",
    "- Our images are only grayscale, but the model assumed colors. We can expect this to have some negative (or at least wasted computation) on our performance.\n",
    "- We needed to resize our images in order for them to fit. We can also expect this to have negative impact on performance."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Bonus Excercise\n",
    "Try improving the model above. Consider the options below\n",
    "- Adding more layers from the ResNet50 model\n",
    "- Adding dropout.\n",
    "- Adding more dense layers/more dense units."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "Layer (type)                        Output Shape                    Param #     \n",
      "================================================================================\n",
      "model_4 (Model)                     (None, 1, 1, 2048)              23587712    \n",
      "________________________________________________________________________________\n",
      "dropout_8 (Dropout)                 (None, 1, 1, 2048)              0           \n",
      "________________________________________________________________________________\n",
      "flatten_6 (Flatten)                 (None, 2048)                    0           \n",
      "________________________________________________________________________________\n",
      "dense_14 (Dense)                    (None, 32)                      65568       \n",
      "________________________________________________________________________________\n",
      "dropout_9 (Dropout)                 (None, 32)                      0           \n",
      "________________________________________________________________________________\n",
      "dense_15 (Dense)                    (None, 10)                      330         \n",
      "================================================================================\n",
      "Total params: 23,653,610\n",
      "Trainable params: 65,898\n",
      "Non-trainable params: 23,587,712\n",
      "________________________________________________________________________________\n",
      "Epoch 10 - loss: 1.413046, acc.: 0.5370833, val loss: 1.29038369655609, val acc.: 0.622500002384186  \r"
     ]
    },
    {
     "ename": "ERROR",
     "evalue": "Error in py_call_impl(callable, dots$args, dots$keywords): TypeError: 'NoneType' object is not callable\n\nDetailed traceback: \n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py\", line 880, in fit\n    validation_steps=validation_steps)\n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/keras/engine/training_arrays.py\", line 329, in model_iteration\n    batch_outs = f(ins_batch)\n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/keras/backend.py\", line 3076, in __call__\n    run_metadata=self.run_metadata)\n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py\", line 1439, in __call__\n    run_metadata_ptr)\n\n",
     "output_type": "error",
     "traceback": [
      "Error in py_call_impl(callable, dots$args, dots$keywords): TypeError: 'NoneType' object is not callable\n\nDetailed traceback: \n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py\", line 880, in fit\n    validation_steps=validation_steps)\n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/keras/engine/training_arrays.py\", line 329, in model_iteration\n    batch_outs = f(ins_batch)\n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/keras/backend.py\", line 3076, in __call__\n    run_metadata=self.run_metadata)\n  File \"/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py\", line 1439, in __call__\n    run_metadata_ptr)\n\nTraceback:\n",
      "1. model %>% fit(x = x_train, y = y_train, validation_split = 0.2, \n .     epochs = 20, batch_size = 128, callbacks = list(Progress$new()))",
      "2. withVisible(eval(quote(`_fseq`(`_lhs`)), env, env))",
      "3. eval(quote(`_fseq`(`_lhs`)), env, env)",
      "4. eval(quote(`_fseq`(`_lhs`)), env, env)",
      "5. `_fseq`(`_lhs`)",
      "6. freduce(value, `_function_list`)",
      "7. withVisible(function_list[[k]](value))",
      "8. function_list[[k]](value)",
      "9. fit(., x = x_train, y = y_train, validation_split = 0.2, epochs = 20, \n .     batch_size = 128, callbacks = list(Progress$new()))",
      "10. fit.keras.engine.training.Model(., x = x_train, y = y_train, \n  .     validation_split = 0.2, epochs = 20, batch_size = 128, callbacks = list(Progress$new()))",
      "11. do.call(object$fit, args)",
      "12. (structure(function (...) \n  . {\n  .     dots <- py_resolve_dots(list(...))\n  .     result <- py_call_impl(callable, dots$args, dots$keywords)\n  .     if (convert) {\n  .         result <- py_to_r(result)\n  .         if (is.null(result)) \n  .             invisible(result)\n  .         else result\n  .     }\n  .     else {\n  .         result\n  .     }\n  . }, class = c(\"python.builtin.method\", \"python.builtin.object\"\n  . ), py_object = <environment>))(batch_size = 128L, epochs = 20L, \n  .     verbose = 1L, callbacks = list(<environment>, <environment>), \n  .     validation_split = 0.2, shuffle = TRUE, class_weight = NULL, \n  .     sample_weight = NULL, initial_epoch = 0L, x = <environment>, \n  .     y = <environment>)",
      "13. py_call_impl(callable, dots$args, dots$keywords)"
     ]
    }
   ],
   "source": [
    "smaller_resnet_50 <- keras_model(inputs = resnet_50_model$input, \n",
    "                                 outputs = get_layer(resnet_50_model, 'activation_48')$output)\n",
    "freeze_weights(smaller_resnet_50)\n",
    "model <- keras_model_sequential() %>%\n",
    "    smaller_resnet_50 %>%\n",
    "    layer_dropout(0.2) %>%\n",
    "    layer_flatten() %>%\n",
    "    layer_dense(units = 32, activation = \"relu\") %>%\n",
    "    layer_dropout(0.2) %>%\n",
    "    layer_dense(units = 10, activation = \"softmax\")\n",
    "\n",
    "cat(summary(model))\n",
    "\n",
    "model %>% compile(\n",
    "    optimizer = optimizer_adam(lr = 0.0001),\n",
    "    loss = \"categorical_crossentropy\",\n",
    "    metric = \"acc\"\n",
    ")\n",
    "\n",
    "history <- model %>% fit(\n",
    "    x = x_train,\n",
    "    y = y_train,\n",
    "    validation_split = 0.2,\n",
    "    epochs = 20,\n",
    "    batch_size = 128,\n",
    "    callbacks=list(Progress$new())\n",
    ")\n",
    "\n",
    "plot(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "R",
   "language": "R",
   "name": "ir"
  },
  "language_info": {
   "codemirror_mode": "r",
   "file_extension": ".r",
   "mimetype": "text/x-r-source",
   "name": "R",
   "pygments_lexer": "r",
   "version": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}