Assignment_2_Babak_Ansari_Convolutional_Neural_Nets_and_Transfer_Learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "view-in-github"
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/babakansari/DeepLearning/blob/master/Assignment_2_Babak_Ansari_Convolutional_Neural_Nets_and_Transfer_Learning.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "NihUzEPbP8JN"
   },
   "source": [
    "# Assignment 2 - Babak Ansari-Jaberi"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Prepration"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Flower images should be downloaded into the current folder\n",
    "  \n",
    "Creating a function to resize the image files in `flower_photos` folder and save them into a new `data`  folder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 259,
   "metadata": {},
   "outputs": [],
   "source": [
    "# required packages that might be missing and settings\n",
    "#!pip3 install PIL  --allow-unverified PIL --allow-all-external\n",
    "%config IPCompleter.greedy=True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Function to convert image sizes and save in a destination folder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 260,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "qKrkPd-nQB-i"
   },
   "outputs": [],
   "source": [
    "import os, sys\n",
    "from PIL import Image\n",
    "\n",
    "# Supress warning and informational messages\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'\n",
    "\n",
    "img_height = 150 \n",
    "img_width = 150\n",
    "\n",
    "# Function to resize a single image\n",
    "def resize_photo(infile, outfile, width, height):   \n",
    "    size = width, height\n",
    "\n",
    "    if not os.path.exists(os.path.dirname(outfile)):\n",
    "        try:\n",
    "            os.makedirs(os.path.dirname(outfile))\n",
    "        except OSError as exc: # Guard against race condition\n",
    "            raise\n",
    "\n",
    "    try:\n",
    "        im = Image.open(infile)\n",
    "        im.thumbnail(size, Image.ANTIALIAS) \n",
    "        im.save(outfile, \"JPEG\")\n",
    "        im = Image.open(outfile)\n",
    "    except IOError:\n",
    "        raise\n",
    "    return im"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Function to add images to Data Frame for normalizing the distribution of flower kinds:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 261,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from os import listdir\n",
    "from os.path import isfile, join\n",
    "from IPython.display import clear_output # Print on the same line to show the progress\n",
    "\n",
    "# Function to populate a data frame of image paths and their kinds\n",
    "def photos_to_dataframe(idx, image_path):\n",
    "    flower_kind = image_path.replace('flower_photos', '').replace('/','')\n",
    "    \n",
    "    onlyfiles = [f for f in listdir(image_path) if isfile(join(image_path, f))]\n",
    "\n",
    "    df = pd.DataFrame(columns=['id', 'kind', 'image', 'path'])\n",
    "    \n",
    "    for imgfile in onlyfiles:\n",
    "        clear_output(wait=True)\n",
    "        print('Processing files in \"' + image_path + '\" folder.\\r\\n'+ imgfile)\n",
    "        df.loc[idx] = [idx, flower_kind, imgfile, image_path]\n",
    "        idx=idx+1\n",
    "    clear_output(wait=True)\n",
    "    print('Processing files in \"' + image_path + '\" folder is finished.')\n",
    "    return df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create a data frame of all images and their flower kinds to be used for preprocessing (e.g. normalizing the flower kinds distribution)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 262,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing files in \"flower_photos/tulips/\" folder is finished.\n"
     ]
    }
   ],
   "source": [
    "daisy_df = photos_to_dataframe(0, 'flower_photos/daisy/')\n",
    "dandelion_df = photos_to_dataframe(len(daisy_df)+1, 'flower_photos/dandelion/')\n",
    "roses_df = photos_to_dataframe(len(dandelion_df)+1, 'flower_photos/roses/')\n",
    "sunflowers_df = photos_to_dataframe(len(roses_df)+1, 'flower_photos/sunflowers/')\n",
    "tulips_df = photos_to_dataframe(len(sunflowers_df)+1, 'flower_photos/tulips/')\n",
    "\n",
    "df_main = daisy_df.append(dandelion_df).append(roses_df).append(sunflowers_df).append(tulips_df).sort_values(by=['kind'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 263,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>kind</th>\n",
       "      <th>image</th>\n",
       "      <th>path</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>daisy</td>\n",
       "      <td>100080576_f52e8ee070_n.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>418</td>\n",
       "      <td>418</td>\n",
       "      <td>daisy</td>\n",
       "      <td>4858518329_7563eb0baa_m.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>419</td>\n",
       "      <td>419</td>\n",
       "      <td>daisy</td>\n",
       "      <td>4861391074_c3e122dab0_m.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>420</td>\n",
       "      <td>420</td>\n",
       "      <td>daisy</td>\n",
       "      <td>4865691548_00319261b8.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>421</td>\n",
       "      <td>421</td>\n",
       "      <td>daisy</td>\n",
       "      <td>488202750_c420cbce61.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      id   kind                        image                  path\n",
       "0      0  daisy   100080576_f52e8ee070_n.jpg  flower_photos/daisy/\n",
       "418  418  daisy  4858518329_7563eb0baa_m.jpg  flower_photos/daisy/\n",
       "419  419  daisy  4861391074_c3e122dab0_m.jpg  flower_photos/daisy/\n",
       "420  420  daisy    4865691548_00319261b8.jpg  flower_photos/daisy/\n",
       "421  421  daisy     488202750_c420cbce61.jpg  flower_photos/daisy/"
      ]
     },
     "execution_count": 263,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_main.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The images have different dimensions. Resize all of them to 150x150.\n",
    "\n",
    "Normalize the flowe kinds folder for analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 272,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>kind</th>\n",
       "      <th>image</th>\n",
       "      <th>path</th>\n",
       "      <th>daisy</th>\n",
       "      <th>dandelion</th>\n",
       "      <th>roses</th>\n",
       "      <th>sunflowers</th>\n",
       "      <th>tulips</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>daisy</td>\n",
       "      <td>100080576_f52e8ee070_n.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>418</td>\n",
       "      <td>418</td>\n",
       "      <td>daisy</td>\n",
       "      <td>4858518329_7563eb0baa_m.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>419</td>\n",
       "      <td>419</td>\n",
       "      <td>daisy</td>\n",
       "      <td>4861391074_c3e122dab0_m.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>420</td>\n",
       "      <td>420</td>\n",
       "      <td>daisy</td>\n",
       "      <td>4865691548_00319261b8.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>421</td>\n",
       "      <td>421</td>\n",
       "      <td>daisy</td>\n",
       "      <td>488202750_c420cbce61.jpg</td>\n",
       "      <td>flower_photos/daisy/</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      id   kind                        image                  path  daisy  \\\n",
       "0      0  daisy   100080576_f52e8ee070_n.jpg  flower_photos/daisy/      1   \n",
       "418  418  daisy  4858518329_7563eb0baa_m.jpg  flower_photos/daisy/      1   \n",
       "419  419  daisy  4861391074_c3e122dab0_m.jpg  flower_photos/daisy/      1   \n",
       "420  420  daisy    4865691548_00319261b8.jpg  flower_photos/daisy/      1   \n",
       "421  421  daisy     488202750_c420cbce61.jpg  flower_photos/daisy/      1   \n",
       "\n",
       "     dandelion  roses  sunflowers  tulips  \n",
       "0            0      0           0       0  \n",
       "418          0      0           0       0  \n",
       "419          0      0           0       0  \n",
       "420          0      0           0       0  \n",
       "421          0      0           0       0  "
      ]
     },
     "execution_count": 272,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# One hot encoding the flower kinds\n",
    "df = pd.concat([df_main,pd.get_dummies(df_main['kind'], drop_first=False)],axis=1)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split images to 75-25% for training and test:\n",
    "\n",
    "Make sure the distribution is random `shuffle=True`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 273,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "train, test = train_test_split(df, test_size=0.25, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Make sure you have the same distribution of flower types between train and test datasets:\n",
    "The distribution of the rows are the same for both train and test.\n",
    "As you could see in bellow statistical summary for train and test set. The average number of all flower kinds are very close. Meaning that each flower kind is distributed evenly in test and training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 274,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>daisy</th>\n",
       "      <th>dandelion</th>\n",
       "      <th>roses</th>\n",
       "      <th>sunflowers</th>\n",
       "      <th>tulips</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>count</td>\n",
       "      <td>2752.000000</td>\n",
       "      <td>2752.000000</td>\n",
       "      <td>2752.000000</td>\n",
       "      <td>2752.000000</td>\n",
       "      <td>2752.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>mean</td>\n",
       "      <td>0.172602</td>\n",
       "      <td>0.248547</td>\n",
       "      <td>0.172965</td>\n",
       "      <td>0.189680</td>\n",
       "      <td>0.216206</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>std</td>\n",
       "      <td>0.377971</td>\n",
       "      <td>0.432249</td>\n",
       "      <td>0.378286</td>\n",
       "      <td>0.392119</td>\n",
       "      <td>0.411731</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>min</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25%</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>50%</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>75%</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>max</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             daisy    dandelion        roses   sunflowers       tulips\n",
       "count  2752.000000  2752.000000  2752.000000  2752.000000  2752.000000\n",
       "mean      0.172602     0.248547     0.172965     0.189680     0.216206\n",
       "std       0.377971     0.432249     0.378286     0.392119     0.411731\n",
       "min       0.000000     0.000000     0.000000     0.000000     0.000000\n",
       "25%       0.000000     0.000000     0.000000     0.000000     0.000000\n",
       "50%       0.000000     0.000000     0.000000     0.000000     0.000000\n",
       "75%       0.000000     0.000000     0.000000     0.000000     0.000000\n",
       "max       1.000000     1.000000     1.000000     1.000000     1.000000"
      ]
     },
     "execution_count": 274,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 275,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>daisy</th>\n",
       "      <th>dandelion</th>\n",
       "      <th>roses</th>\n",
       "      <th>sunflowers</th>\n",
       "      <th>tulips</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>count</td>\n",
       "      <td>918.000000</td>\n",
       "      <td>918.000000</td>\n",
       "      <td>918.000000</td>\n",
       "      <td>918.000000</td>\n",
       "      <td>918.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>mean</td>\n",
       "      <td>0.172113</td>\n",
       "      <td>0.233115</td>\n",
       "      <td>0.179739</td>\n",
       "      <td>0.192810</td>\n",
       "      <td>0.222222</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>std</td>\n",
       "      <td>0.377685</td>\n",
       "      <td>0.423046</td>\n",
       "      <td>0.384179</td>\n",
       "      <td>0.394721</td>\n",
       "      <td>0.415966</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>min</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25%</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>50%</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>75%</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>max</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            daisy   dandelion       roses  sunflowers      tulips\n",
       "count  918.000000  918.000000  918.000000  918.000000  918.000000\n",
       "mean     0.172113    0.233115    0.179739    0.192810    0.222222\n",
       "std      0.377685    0.423046    0.384179    0.394721    0.415966\n",
       "min      0.000000    0.000000    0.000000    0.000000    0.000000\n",
       "25%      0.000000    0.000000    0.000000    0.000000    0.000000\n",
       "50%      0.000000    0.000000    0.000000    0.000000    0.000000\n",
       "75%      0.000000    0.000000    0.000000    0.000000    0.000000\n",
       "max      1.000000    1.000000    1.000000    1.000000    1.000000"
      ]
     },
     "execution_count": 275,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The mean and std are almost the same in train and test sets as you could see in above summary.\n",
    "\n",
    "Now creating data frames with total flower kinds in train and test to analyse the total flower kinds in tran and test set:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 276,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>kind</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>daisy</td>\n",
       "      <td>633</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>dandelion</td>\n",
       "      <td>898</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>roses</td>\n",
       "      <td>641</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>sunflowers</td>\n",
       "      <td>699</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>tulips</td>\n",
       "      <td>799</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         kind  count\n",
       "0       daisy    633\n",
       "1   dandelion    898\n",
       "2       roses    641\n",
       "3  sunflowers    699\n",
       "4      tulips    799"
      ]
     },
     "execution_count": 276,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "df_kinds = pd.DataFrame({'count' : df.groupby( [ \"kind\" ] ).size()}).reset_index()\n",
    "train_kinds = pd.DataFrame({'count' : train.groupby( [ \"kind\" ] ).size()}).reset_index()\n",
    "test_kinds = pd.DataFrame({'count' : test.groupby( [ \"kind\" ] ).size()}).reset_index()\n",
    "df_kinds"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Normalize the counts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 277,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import preprocessing\n",
    "\n",
    "\n",
    "df_kinds[['count']] = preprocessing.minmax_scale(df_kinds[['count']])\n",
    "train_kinds[['count']] = preprocessing.minmax_scale(train_kinds[['count']])\n",
    "test_kinds[['count']] = preprocessing.minmax_scale(test_kinds[['count']])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following graph bellow is showing that the distribution of flower kinds are normal between Blue (Entire set), Red (Train set) and (Green set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 278,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x233df6950>"
      ]
     },
     "execution_count": 278,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.distplot(df_kinds['count'], color=\"Blue\", label=\"All\")\n",
    "sns.distplot(train_kinds['count'], color=\"Red\")\n",
    "sns.distplot(test_kinds['count'], color=\"Green\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following bar charts are also showing the distribution of flower kinds are the same (With different scaling) on test and training data sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 279,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Text(0, 0, 'daisy'),\n",
       " Text(0, 0, 'dandelion'),\n",
       " Text(0, 0, 'roses'),\n",
       " Text(0, 0, 'sunflowers'),\n",
       " Text(0, 0, 'tulips')]"
      ]
     },
     "execution_count": 279,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1202.4x504 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "fig, (ax1, ax2, ax3) = plt.subplots(ncols=3, sharey=True, figsize=(16.7, 7))\n",
    "\n",
    "a1 = sns.barplot(x=\"kind\", y=\"count\", hue=\"kind\", data=df_kinds, palette=\"Blues\", ax=ax1)\n",
    "a2 = sns.barplot(x=\"kind\", y=\"count\", hue=\"kind\", data=train_kinds, palette=\"Reds\", ax=ax2)\n",
    "a3 = sns.barplot(x=\"kind\", y=\"count\", hue=\"kind\", data=test_kinds, palette=\"Greens\", ax=ax3)\n",
    "\n",
    "a1.set_xticklabels(a1.get_xticklabels(), rotation=45)\n",
    "a2.set_xticklabels(a2.get_xticklabels(), rotation=45)\n",
    "a3.set_xticklabels(a3.get_xticklabels(), rotation=45)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Now we know our dataset is correct so we could create the trainable folder for Keras.\n",
    "\n",
    "We could create a trainable folder structure for `Keras` now that we know our training and test set are having same flower type distribution.\n",
    "\n",
    "For that we could loop throught both train and test data frames and create `data\\train\\{flower kind}` and `data\\validate\\{flower kind}` folders respectively."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 304,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_files_for_training(df, train_test_name):\n",
    "    for index, row in df.iterrows():\n",
    "        source_loc = row['path']+row['image']\n",
    "        dest_loc = 'data/' + train_test_name + '/' + row['kind'] + '/' + row['image']\n",
    "        img = resize_photo(source_loc, dest_loc, 150, 150)\n",
    "        clear_output(wait=True)\n",
    "        print('Create flower image in '+ dest_loc + ' folder.\\r\\n')\n",
    "    print('Creating images for '+train_test_name+' in ' + 'data/' + train_test_name  + ' folder is done.')\n",
    "    return 'data/' + train_test_name"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create the train and test folders:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 305,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Create flower image in data/validate/tulips/15632065904_0d9caf174b.jpg folder.\r\n",
      "\n",
      "Creating images for validate in data/validate folder is done.\n",
      "The training folder is: data/train and validation folder is: data/validate\n"
     ]
    }
   ],
   "source": [
    "train_folder = create_files_for_training(train, 'train')\n",
    "validate_folder = create_files_for_training(test, 'validate')\n",
    "\n",
    "print('The training folder is: '+ train_folder + ' and validation folder is: '+validate_folder)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Use a VGG16 model (pre-trained on ImageNet)\n",
    "Use VGG16 class has a parameter called `include_top` which could remove the top layers (fully connected layers) when is set `False`. \n",
    "Trying to remove the layer using `layers.pop()` function is not as straight so use this feature to remove the top layers:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 319,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"vgg16\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_23 (InputLayer)        (None, 150, 150, 3)       0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0         \n",
      "=================================================================\n",
      "Total params: 14,714,688\n",
      "Trainable params: 14,714,688\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.applications.vgg16 import VGG16\n",
    "from keras.preprocessing import image\n",
    "from keras.applications.vgg16 import preprocess_input\n",
    "from keras.layers import Input, Flatten, Dense\n",
    "from keras.models import Model\n",
    "import numpy as np\n",
    "\n",
    "#Get back the convolutional part of a VGG network trained on ImageNet\n",
    "input_tensor = Input(shape=(150,150,3))\n",
    "model_vgg16_conv = VGG16(weights='imagenet', include_top=False, input_tensor=input_tensor)\n",
    "model_vgg16_conv.summary()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Add fully connected layers\n",
    "One with 256 nodes using ‘relu’ activation and output layer with 5 nodes and ‘softmax’ activation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 320,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "last = model_vgg16_conv.output\n",
    "\n",
    "x = Flatten(name='flatten_layer')(last)\n",
    "x = Dense(256, activation='relu')(x)\n",
    "preds = Dense(5, activation='softmax')(x)\n",
    "\n",
    "new_model = Model(model_vgg16_conv.input, preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 321,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_14\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_23 (InputLayer)        (None, 150, 150, 3)       0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0         \n",
      "_________________________________________________________________\n",
      "flatten_layer (Flatten)      (None, 8192)              0         \n",
      "_________________________________________________________________\n",
      "dense_54 (Dense)             (None, 256)               2097408   \n",
      "_________________________________________________________________\n",
      "dense_55 (Dense)             (None, 5)                 1285      \n",
      "=================================================================\n",
      "Total params: 16,813,381\n",
      "Trainable params: 16,813,381\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "new_model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Freez all VGG16 layers\n",
    "Train (fine-tune) and evaluate the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 326,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The input_23 layer freezing status is False\n",
      "The block1_conv1 layer freezing status is True\n",
      "The block1_conv2 layer freezing status is True\n",
      "The block1_pool layer freezing status is True\n",
      "The block2_conv1 layer freezing status is True\n",
      "The block2_conv2 layer freezing status is True\n",
      "The block2_pool layer freezing status is True\n",
      "The block3_conv1 layer freezing status is True\n",
      "The block3_conv2 layer freezing status is True\n",
      "The block3_conv3 layer freezing status is True\n",
      "The block3_pool layer freezing status is True\n",
      "The block4_conv1 layer freezing status is True\n",
      "The block4_conv2 layer freezing status is True\n",
      "The block4_conv3 layer freezing status is True\n",
      "The block4_pool layer freezing status is True\n",
      "The block5_conv1 layer freezing status is True\n",
      "The block5_conv2 layer freezing status is True\n",
      "The block5_conv3 layer freezing status is True\n",
      "The block5_pool layer freezing status is True\n",
      "The flatten_layer layer freezing status is True\n",
      "The dense_54 layer freezing status is True\n",
      "The dense_55 layer freezing status is True\n"
     ]
    }
   ],
   "source": [
    "# This function displays the trainability of the layers\n",
    "def print_layers_trainability(model):\n",
    "    for layer in model.layers:\n",
    "        print('The ' + layer.name + ' layer freezing status is ' + str(layer.trainable))\n",
    "\n",
    "print_layers_trainability(new_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 327,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Freezing input_23 layer\n",
      "Freezing block1_conv1 layer\n",
      "Freezing block1_conv2 layer\n",
      "Freezing block1_pool layer\n",
      "Freezing block2_conv1 layer\n",
      "Freezing block2_conv2 layer\n",
      "Freezing block2_pool layer\n",
      "Freezing block3_conv1 layer\n",
      "Freezing block3_conv2 layer\n",
      "Freezing block3_conv3 layer\n",
      "Freezing block3_pool layer\n",
      "Freezing block4_conv1 layer\n",
      "Freezing block4_conv2 layer\n",
      "Freezing block4_conv3 layer\n",
      "Freezing block4_pool layer\n",
      "Freezing block5_conv1 layer\n",
      "Freezing block5_conv2 layer\n",
      "Freezing block5_conv3 layer\n",
      "Freezing block5_pool layer\n",
      "The input_23 layer freezing status is False\n",
      "The block1_conv1 layer freezing status is False\n",
      "The block1_conv2 layer freezing status is False\n",
      "The block1_pool layer freezing status is False\n",
      "The block2_conv1 layer freezing status is False\n",
      "The block2_conv2 layer freezing status is False\n",
      "The block2_pool layer freezing status is False\n",
      "The block3_conv1 layer freezing status is False\n",
      "The block3_conv2 layer freezing status is False\n",
      "The block3_conv3 layer freezing status is False\n",
      "The block3_pool layer freezing status is False\n",
      "The block4_conv1 layer freezing status is False\n",
      "The block4_conv2 layer freezing status is False\n",
      "The block4_conv3 layer freezing status is False\n",
      "The block4_pool layer freezing status is False\n",
      "The block5_conv1 layer freezing status is False\n",
      "The block5_conv2 layer freezing status is False\n",
      "The block5_conv3 layer freezing status is False\n",
      "The block5_pool layer freezing status is False\n",
      "The flatten_layer layer freezing status is True\n",
      "The dense_54 layer freezing status is True\n",
      "The dense_55 layer freezing status is True\n"
     ]
    }
   ],
   "source": [
    "# LOCK THE TOP CONV LAYERS\n",
    "for layer in new_model.layers:\n",
    "    if layer.name == 'flatten_layer':\n",
    "        break;\n",
    "    print('Freezing ' + layer.name + ' layer')\n",
    "    layer.trainable = False\n",
    "    \n",
    "print_layers_trainability(new_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 324,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'data/validate'"
      ]
     },
     "execution_count": 324,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Train folder: ' + train_folder + '\\r\\n')\n",
    "print('Validation folder:' + validate_folder + '\\r\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 325,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 3440 images belonging to 5 classes.\n",
      "Found 1606 images belonging to 5 classes.\n",
      "Epoch 1/3\n",
      "108/108 [==============================] - 936s 9s/step - loss: 1.2661 - accuracy: 0.5047 - val_loss: 0.5446 - val_accuracy: 0.7092\n",
      "Epoch 2/3\n",
      "108/108 [==============================] - 910s 8s/step - loss: 0.6311 - accuracy: 0.7683 - val_loss: 0.1868 - val_accuracy: 0.7983\n",
      "Epoch 3/3\n",
      "108/108 [==============================] - 988s 9s/step - loss: 0.4745 - accuracy: 0.8297 - val_loss: 0.1037 - val_accuracy: 0.8306\n"
     ]
    }
   ],
   "source": [
    "from keras import applications, optimizers\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dropout, Flatten, Dense, Input\n",
    "import keras\n",
    "\n",
    "batch_size = 32\n",
    "epochs = 3\n",
    "# COMPILE THE MODEL - Use categorical crossentropy because we have four classes of flower kinds\n",
    "new_model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=optimizers.SGD(lr=1e-4, momentum=0.9),\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# CREATE THE IMAGE GENERATORS\n",
    "train_datagen = ImageDataGenerator(rescale=1./255,\n",
    "                                   shear_range=0.2,\n",
    "                                   zoom_range=0.2,\n",
    "                                   horizontal_flip=True)\n",
    "validation_datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "                        train_folder,\n",
    "                        target_size=(img_height,img_width),\n",
    "                        batch_size=batch_size)\n",
    "\n",
    "validation_generator = validation_datagen.flow_from_directory(\n",
    "                            validate_folder,\n",
    "                            target_size=(img_height,img_width),\n",
    "                            batch_size=batch_size)\n",
    "\n",
    "\n",
    "#  FIT THE MODEL\n",
    "history = new_model.fit_generator(\n",
    "    train_generator,\n",
    "    epochs=epochs,\n",
    "    validation_data=validation_generator)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Save the model for future use:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 317,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_model.save('flower_category1.model')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Experiments\n",
    "- First try with 5 epoch the model had `0.89` percent accuracy and `0.36` percent loss. But reduce the epochs to 3 to have faster training performance.\n",
    "- Second experiment is with 3 epochs. The model had `0.84` percent accuracy and `0.10` percent loss.\n",
    "-  The Loss and Accuracy plot shows the model is improving in each epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 344,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "### plt.style.use(\"ggplot\")\n",
    "def plot_model(training_history):\n",
    "    plt.figure()\n",
    "    plt.plot(np.arange(0, 3), training_history.history[\"loss\"], label=\"train_loss\")\n",
    "    plt.plot(np.arange(0, 3), training_history.history[\"val_loss\"], label=\"val_loss\")\n",
    "    plt.plot(np.arange(0, 3), training_history.history[\"accuracy\"], label=\"train_acc\")\n",
    "    plt.plot(np.arange(0, 3), training_history.history[\"val_accuracy\"], label=\"val_acc\")\n",
    "    plt.title(\"Training Loss and Accuracy on Dataset\")\n",
    "    plt.xlabel(\"Epoch #\")\n",
    "    plt.ylabel(\"Loss/Accuracy\")\n",
    "    plt.legend(loc=\"lower left\")\n",
    "    \n",
    "plot_model(history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Unfreeze the last block of VGG16 (block5)\n",
    "\n",
    "Re-train and evaluate the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 340,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The input_17 layer freezing status is False\n",
      "The block1_conv1 layer freezing status is False\n",
      "The block1_conv2 layer freezing status is False\n",
      "The block1_pool layer freezing status is False\n",
      "The block2_conv1 layer freezing status is False\n",
      "The block2_conv2 layer freezing status is False\n",
      "The block2_pool layer freezing status is False\n",
      "The block3_conv1 layer freezing status is False\n",
      "The block3_conv2 layer freezing status is False\n",
      "The block3_conv3 layer freezing status is False\n",
      "The block3_pool layer freezing status is False\n",
      "The block4_conv1 layer freezing status is False\n",
      "The block4_conv2 layer freezing status is False\n",
      "The block4_conv3 layer freezing status is False\n",
      "The block4_pool layer freezing status is False\n",
      "The block5_conv1 layer freezing status is True\n",
      "The block5_conv2 layer freezing status is True\n",
      "The block5_conv3 layer freezing status is True\n",
      "The block5_pool layer freezing status is True\n",
      "The flatten_layer layer freezing status is True\n",
      "The dense_40 layer freezing status is True\n",
      "The dense_41 layer freezing status is True\n"
     ]
    }
   ],
   "source": [
    "\n",
    "model.get_layer('block5_conv1').trainable = True\n",
    "model.get_layer('block5_conv2').trainable = True\n",
    "model.get_layer('block5_conv3').trainable = True\n",
    "model.get_layer('block5_pool').trainable = True\n",
    "\n",
    "print_layers_trainability(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 341,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/3\n",
      "108/108 [==============================] - 949s 9s/step - loss: 0.4142 - accuracy: 0.8468 - val_loss: 0.2434 - val_accuracy: 0.8263\n",
      "Epoch 2/3\n",
      "108/108 [==============================] - 905s 8s/step - loss: 0.3648 - accuracy: 0.8666 - val_loss: 0.0970 - val_accuracy: 0.8599\n",
      "Epoch 3/3\n",
      "108/108 [==============================] - 956s 9s/step - loss: 0.3334 - accuracy: 0.8791 - val_loss: 0.3396 - val_accuracy: 0.8780\n"
     ]
    }
   ],
   "source": [
    "#  FIT THE MODEL\n",
    "history2 = new_model.fit_generator(\n",
    "    train_generator,\n",
    "    epochs=epochs,\n",
    "    validation_data=validation_generator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_model.save('flower_category2.model')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Experiments\n",
    "The experiment for unfreeze the last block of VGG16 with 3 epochs:\n",
    "- The model had `0.88` percent accuracy and `0.34` percent loss.\n",
    "- It seems adding the last block layer improved the performance eventhough the loss was more\n",
    "- The Loss and Accuracy plot shows the model is improving in each epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 345,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "plot_model(history2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Unfreeze all the layers and try again. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 346,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Un-freezing input_23 layer\n",
      "Un-freezing block1_conv1 layer\n",
      "Un-freezing block1_conv2 layer\n",
      "Un-freezing block1_pool layer\n",
      "Un-freezing block2_conv1 layer\n",
      "Un-freezing block2_conv2 layer\n",
      "Un-freezing block2_pool layer\n",
      "Un-freezing block3_conv1 layer\n",
      "Un-freezing block3_conv2 layer\n",
      "Un-freezing block3_conv3 layer\n",
      "Un-freezing block3_pool layer\n",
      "Un-freezing block4_conv1 layer\n",
      "Un-freezing block4_conv2 layer\n",
      "Un-freezing block4_conv3 layer\n",
      "Un-freezing block4_pool layer\n",
      "Un-freezing block5_conv1 layer\n",
      "Un-freezing block5_conv2 layer\n",
      "Un-freezing block5_conv3 layer\n",
      "Un-freezing block5_pool layer\n",
      "Un-freezing flatten_layer layer\n",
      "Un-freezing dense_54 layer\n",
      "Un-freezing dense_55 layer\n",
      "The input_23 layer freezing status is True\n",
      "The block1_conv1 layer freezing status is True\n",
      "The block1_conv2 layer freezing status is True\n",
      "The block1_pool layer freezing status is True\n",
      "The block2_conv1 layer freezing status is True\n",
      "The block2_conv2 layer freezing status is True\n",
      "The block2_pool layer freezing status is True\n",
      "The block3_conv1 layer freezing status is True\n",
      "The block3_conv2 layer freezing status is True\n",
      "The block3_conv3 layer freezing status is True\n",
      "The block3_pool layer freezing status is True\n",
      "The block4_conv1 layer freezing status is True\n",
      "The block4_conv2 layer freezing status is True\n",
      "The block4_conv3 layer freezing status is True\n",
      "The block4_pool layer freezing status is True\n",
      "The block5_conv1 layer freezing status is True\n",
      "The block5_conv2 layer freezing status is True\n",
      "The block5_conv3 layer freezing status is True\n",
      "The block5_pool layer freezing status is True\n",
      "The flatten_layer layer freezing status is True\n",
      "The dense_54 layer freezing status is True\n",
      "The dense_55 layer freezing status is True\n"
     ]
    }
   ],
   "source": [
    "# Unfreez all layers\n",
    "for layer in new_model.layers:\n",
    "    print('Un-freezing ' + layer.name + ' layer')\n",
    "    layer.trainable = True\n",
    "    \n",
    "print_layers_trainability(new_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 347,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/3\n",
      "108/108 [==============================] - 890s 8s/step - loss: 0.2764 - accuracy: 0.9032 - val_loss: 0.1988 - val_accuracy: 0.9029\n",
      "Epoch 2/3\n",
      "108/108 [==============================] - 887s 8s/step - loss: 0.2497 - accuracy: 0.9119 - val_loss: 0.0399 - val_accuracy: 0.9085\n",
      "Epoch 3/3\n",
      "108/108 [==============================] - 949s 9s/step - loss: 0.2248 - accuracy: 0.9166 - val_loss: 0.1996 - val_accuracy: 0.9303\n"
     ]
    }
   ],
   "source": [
    "#  FIT THE MODEL\n",
    "history3 = new_model.fit_generator(\n",
    "    train_generator,\n",
    "    epochs=epochs,\n",
    "    validation_data=validation_generator)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Experiments\n",
    "The experiment for unfreeze all layers epochs:\n",
    "- The model had `0.93` percent accuracy and `0.20` percent loss.\n",
    "- The performance significantly increase after adding all layers to the training\n",
    "- The Loss and Accuracy plot shows the model is improving in each epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 348,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_model(history3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compare the accuracy you got in both cases . \n",
    "Which one is better and why?\n",
    "\n",
    "As stated above, it seems the model with layers participating on training performed the best but most likely less efficient for more data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "include_colab_link": true,
   "name": "Assignment 2 - Babak Ansari - Convolutional Neural Nets and Transfer Learning",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}