Wafer Defect Classification CNN model_NC.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This code is written to ingest WM811K dataset and get trained on defect patterns and be able to distinguish and group wafers into two groups - defect pattern vs no pattern. Data was resized to 60x60. Runs existing data defect patterns without any additional data added to rebalance class. \n",
    "\n",
    "Sources:\n",
    "1. Wafer defect classification by deep learning code: Ashadullah Shawon: https://www.kaggle.com/shawon10/wafer-defect-classification-by-deep-learning\n",
    "2. Defect detection in wafer maps: Paul Bassaler: https://www.kaggle.com/paulbassaler/defect-detection-in-wafer-bin-maps\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: tensorflow in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (2.8.0)\n",
      "Requirement already satisfied: libclang>=9.0.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (13.0.0)\n",
      "Requirement already satisfied: flatbuffers>=1.12 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (2.0)\n",
      "Requirement already satisfied: protobuf>=3.9.2 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (3.19.4)\n",
      "Requirement already satisfied: h5py>=2.9.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (2.10.0)\n",
      "Requirement already satisfied: astunparse>=1.6.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.6.3)\n",
      "Requirement already satisfied: tf-estimator-nightly==2.8.0.dev2021122109 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (2.8.0.dev2021122109)\n",
      "Requirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (0.24.0)\n",
      "Requirement already satisfied: termcolor>=1.1.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.1.0)\n",
      "Requirement already satisfied: absl-py>=0.4.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.0.0)\n",
      "Requirement already satisfied: gast>=0.2.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (0.5.3)\n",
      "Requirement already satisfied: numpy>=1.20 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.22.2)\n",
      "Requirement already satisfied: opt-einsum>=2.3.2 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (3.3.0)\n",
      "Requirement already satisfied: typing-extensions>=3.6.6 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (3.7.4.2)\n",
      "Requirement already satisfied: wrapt>=1.11.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.11.2)\n",
      "Requirement already satisfied: keras-preprocessing>=1.1.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.1.2)\n",
      "Requirement already satisfied: google-pasta>=0.1.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (0.2.0)\n",
      "Requirement already satisfied: setuptools in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (49.2.0.post20200714)\n",
      "Requirement already satisfied: six>=1.12.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.15.0)\n",
      "Requirement already satisfied: tensorboard<2.9,>=2.8 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (2.8.0)\n",
      "Requirement already satisfied: grpcio<2.0,>=1.24.3 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (1.43.0)\n",
      "Requirement already satisfied: keras<2.9,>=2.8.0rc0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorflow) (2.8.0)\n",
      "Requirement already satisfied: wheel<1.0,>=0.23.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from astunparse>=1.6.0->tensorflow) (0.34.2)\n",
      "Requirement already satisfied: werkzeug>=0.11.15 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (1.0.1)\n",
      "Requirement already satisfied: requests<3,>=2.21.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (2.24.0)\n",
      "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (0.4.6)\n",
      "Requirement already satisfied: markdown>=2.6.8 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (3.3.6)\n",
      "Requirement already satisfied: google-auth<3,>=1.6.3 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (2.6.0)\n",
      "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (1.8.1)\n",
      "Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from tensorboard<2.9,>=2.8->tensorflow) (0.6.1)\n",
      "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (1.25.9)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (2020.6.20)\n",
      "Requirement already satisfied: chardet<4,>=3.0.2 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (3.0.4)\n",
      "Requirement already satisfied: idna<3,>=2.5 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (2.10)\n",
      "Requirement already satisfied: requests-oauthlib>=0.7.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.9,>=2.8->tensorflow) (1.3.1)\n",
      "Requirement already satisfied: importlib-metadata>=4.4; python_version < \"3.10\" in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from markdown>=2.6.8->tensorboard<2.9,>=2.8->tensorflow) (4.10.1)\n",
      "Requirement already satisfied: cachetools<6.0,>=2.0.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (5.0.0)\n",
      "Requirement already satisfied: rsa<5,>=3.1.4; python_version >= \"3.6\" in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (4.8)\n",
      "Requirement already satisfied: pyasn1-modules>=0.2.1 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (0.2.8)\n",
      "Requirement already satisfied: oauthlib>=3.0.0 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.9,>=2.8->tensorflow) (3.2.0)\n",
      "Requirement already satisfied: zipp>=0.5 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from importlib-metadata>=4.4; python_version < \"3.10\"->markdown>=2.6.8->tensorboard<2.9,>=2.8->tensorflow) (3.1.0)\n",
      "Requirement already satisfied: pyasn1>=0.1.3 in /Users/nagachandrasekaran/opt/anaconda3/lib/python3.8/site-packages (from rsa<5,>=3.1.4; python_version >= \"3.6\"->google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (0.4.8)\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "from os.path import join\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import random\n",
    "\n",
    "!pip install tensorflow\n",
    "import tensorflow as tf\n",
    "import keras\n",
    "from keras import layers, Input, models\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from keras.wrappers.scikit_learn import KerasClassifier \n",
    "from sklearn.model_selection import KFold \n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Import modeling libraries.\n",
    "import sklearn as sk\n",
    "#sklearn.__version__\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.datasets import fetch_openml\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.naive_bayes import BernoulliNB\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.metrics import plot_confusion_matrix\n",
    "from sklearn import metrics\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from skimage.transform import resize as sk_resize\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 811457 entries, 0 to 811456\n",
      "Data columns (total 6 columns):\n",
      " #   Column          Non-Null Count   Dtype  \n",
      "---  ------          --------------   -----  \n",
      " 0   waferMap        811457 non-null  object \n",
      " 1   dieSize         811457 non-null  float64\n",
      " 2   lotName         811457 non-null  object \n",
      " 3   waferIndex      811457 non-null  float64\n",
      " 4   trianTestLabel  811457 non-null  object \n",
      " 5   failureType     811457 non-null  object \n",
      "dtypes: float64(2), object(4)\n",
      "memory usage: 37.1+ MB\n"
     ]
    },
    {
     "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>waferMap</th>\n",
       "      <th>dieSize</th>\n",
       "      <th>lotName</th>\n",
       "      <th>waferIndex</th>\n",
       "      <th>trianTestLabel</th>\n",
       "      <th>failureType</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>506082</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 2,...</td>\n",
       "      <td>710.0</td>\n",
       "      <td>lot31449</td>\n",
       "      <td>2.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>784165</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...</td>\n",
       "      <td>1414.0</td>\n",
       "      <td>lot46434</td>\n",
       "      <td>7.0</td>\n",
       "      <td>[[Test]]</td>\n",
       "      <td>[[none]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>121849</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...</td>\n",
       "      <td>776.0</td>\n",
       "      <td>lot8137</td>\n",
       "      <td>6.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>523097</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1,...</td>\n",
       "      <td>515.0</td>\n",
       "      <td>lot32723</td>\n",
       "      <td>14.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>784238</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...</td>\n",
       "      <td>1089.0</td>\n",
       "      <td>lot46437</td>\n",
       "      <td>21.0</td>\n",
       "      <td>[[Test]]</td>\n",
       "      <td>[[none]]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 waferMap  dieSize   lotName  \\\n",
       "506082  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 2,...    710.0  lot31449   \n",
       "784165  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...   1414.0  lot46434   \n",
       "121849  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...    776.0   lot8137   \n",
       "523097  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1,...    515.0  lot32723   \n",
       "784238  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...   1089.0  lot46437   \n",
       "\n",
       "        waferIndex trianTestLabel failureType  \n",
       "506082         2.0             []          []  \n",
       "784165         7.0       [[Test]]    [[none]]  \n",
       "121849         6.0             []          []  \n",
       "523097        14.0             []          []  \n",
       "784238        21.0       [[Test]]    [[none]]  "
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Read the full data file that is stored locally. Look at information and data details\n",
    "\n",
    "df=pd.read_pickle(\"LSWMD.pkl\")\n",
    "df.info()\n",
    "df.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of unique die size values\n",
      " 710.0     66961\n",
      "515.0     45725\n",
      "712.0     42572\n",
      "1513.0    39322\n",
      "776.0     29513\n",
      "Name: dieSize, dtype: int64\n",
      "\n",
      "Sample of unique dies dize values [1683. 2460.  533. ...  551.  554.  550.]\n",
      "Dies size value that occurs the most 710.0\n"
     ]
    }
   ],
   "source": [
    "# From the dataset we have wafers with different die size and count. Look at details.\n",
    "print(\"Number of unique die size values\\n\", df['dieSize'].value_counts().head())\n",
    "print(\"\\nSample of unique dies dize values\", pd.unique(df['dieSize']))\n",
    "print(\"Dies size value that occurs the most\", df['dieSize'].value_counts().idxmax())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build function to resize images\n",
    "\n",
    "x_dim = 60\n",
    "y_dim = 60\n",
    "\n",
    "def resize_wafers(row):\n",
    "    base_map = row['waferMap']\n",
    "    resize_map = sk_resize(base_map*1024,[x_dim,y_dim], anti_aliasing=True)\n",
    "    return resize_map\n",
    "\n",
    "df['resize_waferMap'] = df.apply (lambda row: resize_wafers(row), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Die size dimension in descending order\n",
      " (32, 29)    108687\n",
      "(25, 27)     64083\n",
      "(49, 39)     39323\n",
      "(26, 26)     30078\n",
      "(30, 34)     29513\n",
      "             ...  \n",
      "(43, 41)         1\n",
      "(49, 67)         1\n",
      "(26, 44)         1\n",
      "(88, 62)         1\n",
      "(54, 64)         1\n",
      "Name: waferMapDim, Length: 632, dtype: int64\n",
      "Die size dimension in descending order\n",
      " (60, 60)    811457\n",
      "Name: resize_waferMapDim, dtype: int64\n"
     ]
    },
    {
     "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>waferMap</th>\n",
       "      <th>dieSize</th>\n",
       "      <th>lotName</th>\n",
       "      <th>waferIndex</th>\n",
       "      <th>trianTestLabel</th>\n",
       "      <th>failureType</th>\n",
       "      <th>resize_waferMap</th>\n",
       "      <th>failureNum</th>\n",
       "      <th>trainTestNum</th>\n",
       "      <th>waferMapDim</th>\n",
       "      <th>resize_waferMapDim</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>436670</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...</td>\n",
       "      <td>3777.0</td>\n",
       "      <td>lot26414</td>\n",
       "      <td>10.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>(63, 76)</td>\n",
       "      <td>(60, 60)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>235429</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...</td>\n",
       "      <td>5633.0</td>\n",
       "      <td>lot14724</td>\n",
       "      <td>19.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>(88, 81)</td>\n",
       "      <td>(60, 60)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>498779</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...</td>\n",
       "      <td>1513.0</td>\n",
       "      <td>lot31048</td>\n",
       "      <td>22.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>(49, 39)</td>\n",
       "      <td>(60, 60)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>534219</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 2,...</td>\n",
       "      <td>710.0</td>\n",
       "      <td>lot33390</td>\n",
       "      <td>3.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>(32, 29)</td>\n",
       "      <td>(60, 60)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>581191</th>\n",
       "      <td>[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 1,...</td>\n",
       "      <td>846.0</td>\n",
       "      <td>lot36292</td>\n",
       "      <td>15.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[]</td>\n",
       "      <td>(33, 33)</td>\n",
       "      <td>(60, 60)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 waferMap  dieSize   lotName  \\\n",
       "436670  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...   3777.0  lot26414   \n",
       "235429  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...   5633.0  lot14724   \n",
       "498779  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...   1513.0  lot31048   \n",
       "534219  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 2,...    710.0  lot33390   \n",
       "581191  [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 1,...    846.0  lot36292   \n",
       "\n",
       "        waferIndex trianTestLabel failureType  \\\n",
       "436670        10.0             []          []   \n",
       "235429        19.0             []          []   \n",
       "498779        22.0             []          []   \n",
       "534219         3.0             []          []   \n",
       "581191        15.0             []          []   \n",
       "\n",
       "                                          resize_waferMap failureNum  \\\n",
       "436670  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...         []   \n",
       "235429  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...         []   \n",
       "498779  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...         []   \n",
       "534219  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...         []   \n",
       "581191  [[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,...         []   \n",
       "\n",
       "       trainTestNum waferMapDim resize_waferMapDim  \n",
       "436670           []    (63, 76)           (60, 60)  \n",
       "235429           []    (88, 81)           (60, 60)  \n",
       "498779           []    (49, 39)           (60, 60)  \n",
       "534219           []    (32, 29)           (60, 60)  \n",
       "581191           []    (33, 33)           (60, 60)  "
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Re-arranging the data. Creating nuerical values for failure type and train/test type\n",
    "# Also measuring the wafer map dimensions and adding a column\n",
    "# waferMap value is a numpy array with 0, 1, 2 values. 0-no data, 1-pure pixel die, 2-defect die\n",
    "# Each die is represented by a pixel value. So, size of this waferMap array is different\n",
    "\n",
    "df['failureNum']=df.failureType\n",
    "df['trainTestNum']=df.trianTestLabel\n",
    "mapping_type={'Center':0,'Donut':1,'Edge-Loc':2,'Edge-Ring':3,'Loc':4,'Random':5,'Scratch':6,'Near-full':7,'none':8}\n",
    "mapping_traintest={'Training':0,'Test':1}\n",
    "df=df.replace({'failureNum':mapping_type, 'trainTestNum':mapping_traintest})\n",
    "\n",
    "# Originally, we did not resize the images and selected images with 26x26\n",
    "# This function was used to find the original wafer image dimensions and sort them\n",
    "def find_dim(x):\n",
    "    dim0=np.size(x,axis=0)\n",
    "    dim1=np.size(x,axis=1)\n",
    "    return dim0,dim1\n",
    "\n",
    "df['waferMapDim']=df.waferMap.apply(find_dim)\n",
    "print(\"Die size dimension in descending order\\n\", df['waferMapDim'].value_counts().sort_values(ascending=False))\n",
    "\n",
    "# With resized images to 60x60, we can get all images to be same and keep sample size high\n",
    "df['resize_waferMapDim']=df.resize_waferMap.apply(find_dim)\n",
    "print(\"Die size dimension in descending order\\n\", df['resize_waferMapDim'].value_counts().sort_values(ascending=False))\n",
    "\n",
    "df.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wafers with label 172950\n",
      "Wafers with defect patterns 25519\n",
      "Wafers without defect patterns 147431\n"
     ]
    }
   ],
   "source": [
    "# Split the database into three buckets: Wafers with labels, Wafers with pattern defect, Wafers without patterns\n",
    "\n",
    "df_withlabel = df[(df['failureNum']>=0) & (df['failureNum']<=8)]\n",
    "df_withlabel = df_withlabel.reset_index()\n",
    "df_withpattern = df[(df['failureNum']>=0) & (df['failureNum']<=7)]\n",
    "df_withpattern = df_withpattern.reset_index()\n",
    "df_nonpattern = df[(df['failureNum']==8)]\n",
    "print(\"Wafers with label\", df_withlabel.shape[0])\n",
    "print(\"Wafers with defect patterns\", df_withpattern.shape[0])\n",
    "print(\"Wafers without defect patterns\", df_nonpattern.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(15, 5))\n",
    "axe=plt.subplot(1,1,1)\n",
    "\n",
    "uni_pattern=np.unique(df_withpattern.failureNum, return_counts=True)\n",
    "labels = ['Center','Donut','Edge-Loc','Edge-Ring','Loc','Random','Scratch','Near-full', 'None']\n",
    "axe.bar(uni_pattern[0],uni_pattern[1], color='blue', align='center', alpha=0.9)\n",
    "axe.set_title(\"number  of failure types in wafer maps\")\n",
    "axe.set_ylabel(\"number of pattern wafers\")\n",
    "axe.set_xticklabels(labels)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(25519, 60, 60) (25519, 1)\n"
     ]
    }
   ],
   "source": [
    "# We need a precise shape for the wafer maps we will use, it has to be a square\n",
    "# Choose a shape and then we have to create a list of images which are in this shape\n",
    "# Create a correspoinding list with their associated failure types\n",
    "\n",
    "# If we use full dataframe df, there are too many unlabeled values\n",
    "# Even if we eliminate unlabeled values, we still have lot of none defects. Remove them\n",
    "# Use df_withpattern only\n",
    "# Choose 26x26 as the image dimension. Creating a smaller df\n",
    "\n",
    "#sub_df = df_withpattern.loc[df_withpattern['waferMapDim'] == (26, 26)]\n",
    "sub_df = df_withpattern\n",
    "\n",
    "Imlist = np.ones((1, 60, 60)) #originally 60 was 26\n",
    "label = []\n",
    "\n",
    "for i in range(len(sub_df)):\n",
    "    # Remove data without any labels\n",
    "    if len(sub_df.iloc[i,:]['failureType']) == 0:\n",
    "        continue\n",
    "    Imlist = np.concatenate((Imlist, sub_df.iloc[i,:]['resize_waferMap'].reshape(1, 60, 60))) #replaced 26 with 60\n",
    "    label.append(sub_df.iloc[i,:]['failureType'][0][0])\n",
    "    \n",
    "#The first elemet in Imlist is just composed with ones, so it isn't intersesting, let's remove it\n",
    "Imlist=Imlist[1:]\n",
    "label=np.array(label).reshape((-1,1)) #Column of labels\n",
    "\n",
    "#We have all the values of the pixels: 0.0 is for the exterior ofthe wafer map\n",
    "#2.0 is for the local impure pixel and 1.0 is for a pure pixel of the wafer map\n",
    "\n",
    "print(np.shape(Imlist),np.shape(label))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label assiociated with wafer 6096 ['Center']\n",
      "label assiociated with wafer 14896 ['Edge-Loc']\n",
      "label assiociated with wafer 3740 ['Edge-Ring']\n",
      "label assiociated with wafer 18449 ['Edge-Loc']\n",
      "label assiociated with wafer 8250 ['Center']\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, axarr = plt.subplots(1,5)\n",
    "\n",
    "for k in range(5):\n",
    "    j = random.randint(0,25519)\n",
    "    axarr[k].imshow(Imlist[j])\n",
    "    print(\"label assiociated with wafer %s\" % j,label[j])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(25519, 60, 60, 1) 25519\n",
      "New_Imlist shape (25519, 60, 60, 3)\n",
      "[[1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]]\n"
     ]
    }
   ],
   "source": [
    "# Imlist is a numpy array with 14366 entries. Each entry has 26 rows and 26 columns\n",
    "# We now turn all this data into a single column of data\n",
    "# Each pixel value of each image is now separate and we can use it separately to process deep learning\n",
    "Imlist_redo=Imlist.reshape((-1,60,60,1)) #Was 26x26 originally\n",
    "print(np.shape(Imlist_redo), len(Imlist_redo))\n",
    "\n",
    "# In order to process deep learning, we have to use data composed by 1 only. In other words we have to\n",
    "# change the data of 0-blank, 1-clean diw, 2- defect die into a data set composed of 1 only. We have to\n",
    "# transform the dataset into a sequence of 1s for deep learning.\n",
    "# for 0.0 we will use [1, 0, 0]\n",
    "# for 1.0 we will use [0, 1, 0]\n",
    "# for 2.0 we will use [0, 0, 1]\n",
    "\n",
    "new_Imlist = np.zeros((len(Imlist_redo), 60, 60, 3)) #was 26x26 instead of 60x60\n",
    "\n",
    "for w in range(len(Imlist_redo)):\n",
    "    for i in range(60):\n",
    "        for j in range(60):\n",
    "            new_Imlist[w, i, j, int(Imlist_redo[w, i, j])] = 1\n",
    "\n",
    "print(\"New_Imlist shape\", np.shape(new_Imlist))\n",
    "print(new_Imlist[0][1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A Convolutional based Encoder-Decoder Neural Network is an encoder-decoder neural network that consists of a encoder neural network and a decoder neural network in which one or both are convolutional neural networks. \n",
    "\n",
    "An encoder is a network that takes the input, and output a feature map/vector/tensor. These feature vector hold the information, the features, that represents the input. The decoder is again a network (usually the same network structure as encoder but in opposite orientation) that takes the feature vector from the encoder, and gives the best closest match to the actual input or intended output.\n",
    "\n",
    "Once trained, the encoder will gives feature vector for input that can be use by decoder to construct the input with the features that matter the most to make the reconstructed input recognizable as the actual input."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Stochastic gradient descent is an iterative learning algorithm that uses a training dataset to update a model.\n",
    "# The batch size is a hyperparameter of gradient descent that controls the number of training samples to work \n",
    "# through before the model’s internal parameters are updated.The number of epochs is a hyperparameter of gradient \n",
    "# descent that controls the number of complete passes through the training dataset.\n",
    "\n",
    "# One epoch is when entire dataset is passed forward and backward through network once\n",
    "# But instead we train the dataset into different batches and pass it through\n",
    "# If 2000 data samples, divide into 500 batch size, it will take 4 iterations to run 1 epoch\n",
    "epoch=10\n",
    "batch_size=4000\n",
    "\n",
    "input_shape = (60, 60, 3) #Original code was 26 instead of 60x60\n",
    "input_tensor = Input(input_shape)\n",
    "\n",
    "'''\n",
    "Set up the encode layer. Relu activation function, in this case, just keep the values we are convoluting, \n",
    "and it let us do the same for the output tensor created.\n",
    "'''\n",
    "encode = layers.Conv2D(64, (3,3), padding='same', activation='relu')(input_tensor)\n",
    "\n",
    "'''\n",
    "Now that the encoder is settled, we have to code the latent vector, which is the last part of the encoder, and \n",
    "it aims to encapsulate the information for all input elements in order to help the decoder make accurate \n",
    "predictions. It acts as the initial hidden state of the decoder part of the model.\n",
    "'''\n",
    "\n",
    "#Only the highest value of a region is kept.Max pooling is keeping the most activated pixels (ones with the highest \n",
    "#values) and discards the rest.\n",
    "\n",
    "latent_vector = layers.MaxPool2D()(encode)\n",
    "encoder = models.Model(input_tensor, latent_vector)\n",
    "\n",
    "'''\n",
    "And now the decoder, and then connecting all the layers, to finally have the predicted output tensor. A stack of \n",
    "several recurrent units where each predicts an outputEach recurrent unit accepts a hidden state from the previous \n",
    "unit and produces and output as well as its own hidden state.\n",
    "'''\n",
    "\n",
    "#It is the transposed convolution layer, so the inverse step as the encoder\n",
    "decode_layer_1 = layers.Conv2DTranspose(64, (3,3), padding='same', activation='relu') \n",
    "\n",
    "#Repeats the rows and columns of the data by size[0] and size[1] respectively.\n",
    "decode_layer_2 = layers.UpSampling2D() \n",
    "output_tensor = layers.Conv2DTranspose(3, (3,3), padding='same', activation='sigmoid')\n",
    "\n",
    "#After having regarding the shape I needed for the start of the decoder, I have to define it.\n",
    "decoder_input = Input((30, 30, 64)) #decoder input was 13, 13, 64\n",
    "decode = decode_layer_1(decoder_input)\n",
    "decode = decode_layer_2(decode)\n",
    "\n",
    "decoder = models.Model(decoder_input, output_tensor(decode))\n",
    "\n",
    "#Now we have all the model, let's prepare the dataset, with normal wafers, and noised wafers, so that we will have some \n",
    "#data to make the machine predict their label\n",
    "\n",
    "encoded_Im = encoder.predict(new_Imlist)\n",
    "noised_encoded_Im = encoded_Im + np.random.normal(loc=0, scale=0.1, size = (len(encoded_Im), 30, 30, 64))\n",
    "# original code was 13, 13, 64 instead of 30, 30, 64 representing 26x26"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Center : 4294\n",
      "Donut : 555\n",
      "Edge-Loc : 5189\n",
      "Edge-Ring : 9680\n",
      "Loc : 3593\n",
      "Near-full : 149\n",
      "Random : 866\n",
      "Scratch : 1193\n",
      "(25519, 60, 60, 3) (25519, 1)\n"
     ]
    }
   ],
   "source": [
    "#Now we can see the shape of our lists, just to now how big is our dataset\n",
    "for l in np.unique(label) :\n",
    "    print('{} : {}'.format(l, len(label[label==l])))\n",
    "print(np.shape(new_Imlist), np.shape(label))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that this classification is not uniform and it creates a biased distribution of defect types in the dataset. We also do not have sufficient data that has labels and defect patterns. Once we remove unlabeled data and the wafers without defect patterns, we have only 877 entries."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20000, 8)\n",
      "(13400, 60, 60, 3) (6600, 60, 60, 3) (13400, 8) (6600, 8)\n"
     ]
    }
   ],
   "source": [
    "#In order to process the test, we need a matrix of labels\n",
    "for i, l in enumerate(np.unique(label)):\n",
    "    label[label==l] = i\n",
    "    \n",
    "label = tf.keras.utils.to_categorical(label)\n",
    "\n",
    "#Now we know this information, we can lower the heavyness of the dataset, because it is too huge\n",
    "new_IMLIST=new_Imlist[0:20000]\n",
    "new_LABEL=label[0:20000]\n",
    "test_Imlist=new_Imlist[20001:25519]\n",
    "test_label=label[20001:25519]\n",
    "print(np.shape(new_LABEL))\n",
    "\n",
    "im_train, im_test, label_train, label_test = train_test_split(new_IMLIST, new_LABEL, test_size=0.33, random_state=2019)\n",
    "print(np.shape(im_train),np.shape(im_test),np.shape(label_train),np.shape(label_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Let's now create the real model we will use for the CNN test\n",
    "def create_model():\n",
    "    input_shape = (60, 60, 3) #was 26x26x3\n",
    "    input_tensor = Input(input_shape)\n",
    "\n",
    "    conv_1 = layers.Conv2D(16, (3,3), activation='relu', padding='same')(input_tensor)\n",
    "    conv_2 = layers.Conv2D(64, (3,3), activation='relu', padding='same')(conv_1)\n",
    "    conv_3 = layers.Conv2D(128, (3,3), activation='relu', padding='same')(conv_2)\n",
    "\n",
    "    flat = layers.Flatten()(conv_3)\n",
    "\n",
    "    dense_1 = layers.Dense(512, activation='relu')(flat)\n",
    "    dense_2 = layers.Dense(128, activation='relu')(dense_1)\n",
    "    output_tensor = layers.Dense(8, activation='softmax')(dense_2)\n",
    "\n",
    "    model = models.Model(input_tensor, output_tensor)\n",
    "    model.compile(optimizer='Adam',\n",
    "                 loss='categorical_crossentropy',\n",
    "                 metrics=['accuracy'])\n",
    "\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "3/3 - 114s - loss: 20.7536 - accuracy: 0.4095 - 114s/epoch - 38s/step\n",
      "Epoch 2/10\n",
      "3/3 - 106s - loss: 9.1430 - accuracy: 0.1725 - 106s/epoch - 35s/step\n",
      "Epoch 3/10\n",
      "3/3 - 112s - loss: 1.8487 - accuracy: 0.4332 - 112s/epoch - 37s/step\n",
      "Epoch 4/10\n",
      "3/3 - 107s - loss: 1.6320 - accuracy: 0.4332 - 107s/epoch - 36s/step\n",
      "Epoch 5/10\n",
      "3/3 - 99s - loss: 1.6058 - accuracy: 0.4332 - 99s/epoch - 33s/step\n",
      "Epoch 6/10\n",
      "3/3 - 96s - loss: 1.5951 - accuracy: 0.4332 - 96s/epoch - 32s/step\n",
      "Epoch 7/10\n",
      "3/3 - 97s - loss: 1.5809 - accuracy: 0.4332 - 97s/epoch - 32s/step\n",
      "Epoch 8/10\n",
      "3/3 - 96s - loss: 1.5843 - accuracy: 0.4332 - 96s/epoch - 32s/step\n",
      "Epoch 9/10\n",
      "3/3 - 98s - loss: 1.5866 - accuracy: 0.4332 - 98s/epoch - 33s/step\n",
      "Epoch 10/10\n",
      "3/3 - 91s - loss: 1.5878 - accuracy: 0.4332 - 91s/epoch - 30s/step\n",
      "2/2 - 12s - loss: 1.6015 - accuracy: 0.4303 - 12s/epoch - 6s/step\n",
      "Epoch 1/10\n",
      "3/3 - 109s - loss: 52.0327 - accuracy: 0.2215 - 109s/epoch - 36s/step\n",
      "Epoch 2/10\n",
      "3/3 - 98s - loss: 9.7338 - accuracy: 0.3268 - 98s/epoch - 33s/step\n",
      "Epoch 3/10\n",
      "3/3 - 97s - loss: 2.2027 - accuracy: 0.1518 - 97s/epoch - 32s/step\n",
      "Epoch 4/10\n",
      "3/3 - 98s - loss: 1.9052 - accuracy: 0.4318 - 98s/epoch - 33s/step\n",
      "Epoch 5/10\n",
      "3/3 - 90s - loss: 1.7022 - accuracy: 0.4318 - 90s/epoch - 30s/step\n",
      "Epoch 6/10\n",
      "3/3 - 90s - loss: 1.6013 - accuracy: 0.4318 - 90s/epoch - 30s/step\n",
      "Epoch 7/10\n",
      "3/3 - 95s - loss: 1.6054 - accuracy: 0.4318 - 95s/epoch - 32s/step\n",
      "Epoch 8/10\n",
      "3/3 - 97s - loss: 1.5878 - accuracy: 0.4318 - 97s/epoch - 32s/step\n",
      "Epoch 9/10\n",
      "3/3 - 92s - loss: 1.5808 - accuracy: 0.4318 - 92s/epoch - 31s/step\n",
      "Epoch 10/10\n",
      "3/3 - 94s - loss: 1.5836 - accuracy: 0.4318 - 94s/epoch - 31s/step\n",
      "2/2 - 12s - loss: 1.5779 - accuracy: 0.4332 - 12s/epoch - 6s/step\n",
      "Epoch 1/10\n",
      "3/3 - 105s - loss: 10.6765 - accuracy: 0.3981 - 105s/epoch - 35s/step\n",
      "Epoch 2/10\n",
      "3/3 - 94s - loss: 3.4534 - accuracy: 0.2123 - 94s/epoch - 31s/step\n",
      "Epoch 3/10\n",
      "3/3 - 93s - loss: 1.7743 - accuracy: 0.4317 - 93s/epoch - 31s/step\n",
      "Epoch 4/10\n",
      "3/3 - 89s - loss: 1.6588 - accuracy: 0.4317 - 89s/epoch - 30s/step\n",
      "Epoch 5/10\n",
      "3/3 - 88s - loss: 1.6032 - accuracy: 0.4317 - 88s/epoch - 29s/step\n",
      "Epoch 6/10\n",
      "3/3 - 88s - loss: 1.5926 - accuracy: 0.4317 - 88s/epoch - 29s/step\n",
      "Epoch 7/10\n",
      "3/3 - 88s - loss: 1.5874 - accuracy: 0.4317 - 88s/epoch - 29s/step\n",
      "Epoch 8/10\n",
      "3/3 - 91s - loss: 1.5834 - accuracy: 0.4317 - 91s/epoch - 30s/step\n",
      "Epoch 9/10\n",
      "3/3 - 95s - loss: 1.5832 - accuracy: 0.4317 - 95s/epoch - 32s/step\n",
      "Epoch 10/10\n",
      "3/3 - 93s - loss: 1.5842 - accuracy: 0.4317 - 93s/epoch - 31s/step\n",
      "2/2 - 12s - loss: 1.5683 - accuracy: 0.4333 - 12s/epoch - 6s/step\n",
      "[0.43026641 0.43317664 0.43327361]\n",
      "training accuracy =  0.4322388867537181\n"
     ]
    }
   ],
   "source": [
    "model = KerasClassifier(build_fn=create_model, epochs=10, batch_size=4000, verbose=2) \n",
    "# 3-Fold Crossvalidation\n",
    "kfold = KFold(n_splits=3, shuffle=True, random_state=2019) \n",
    "results = cross_val_score(model, im_train, label_train, cv=kfold)\n",
    "print(results)\n",
    "print(\"training accuracy = \", np.mean(results))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "4/4 - 184s - loss: 7.7226 - accuracy: 0.1912 - val_loss: 1.8758 - val_accuracy: 0.1827 - 184s/epoch - 46s/step\n",
      "Epoch 2/10\n",
      "4/4 - 160s - loss: 1.8507 - accuracy: 0.3542 - val_loss: 1.7383 - val_accuracy: 0.4256 - 160s/epoch - 40s/step\n",
      "Epoch 3/10\n",
      "4/4 - 155s - loss: 1.7113 - accuracy: 0.4322 - val_loss: 1.6368 - val_accuracy: 0.4256 - 155s/epoch - 39s/step\n",
      "Epoch 4/10\n",
      "4/4 - 156s - loss: 1.6250 - accuracy: 0.4322 - val_loss: 1.6047 - val_accuracy: 0.4256 - 156s/epoch - 39s/step\n",
      "Epoch 5/10\n",
      "4/4 - 156s - loss: 1.5957 - accuracy: 0.4322 - val_loss: 1.5945 - val_accuracy: 0.4256 - 156s/epoch - 39s/step\n",
      "Epoch 6/10\n",
      "4/4 - 160s - loss: 1.5806 - accuracy: 0.4322 - val_loss: 1.5874 - val_accuracy: 0.4256 - 160s/epoch - 40s/step\n",
      "Epoch 7/10\n",
      "4/4 - 161s - loss: 1.5769 - accuracy: 0.4322 - val_loss: 1.5881 - val_accuracy: 0.4256 - 161s/epoch - 40s/step\n",
      "Epoch 8/10\n",
      "4/4 - 160s - loss: 1.5769 - accuracy: 0.4322 - val_loss: 1.5883 - val_accuracy: 0.4256 - 160s/epoch - 40s/step\n",
      "Epoch 9/10\n",
      "4/4 - 181s - loss: 1.5749 - accuracy: 0.4322 - val_loss: 1.5883 - val_accuracy: 0.4256 - 181s/epoch - 45s/step\n",
      "Epoch 10/10\n",
      "4/4 - 177s - loss: 1.5755 - accuracy: 0.4322 - val_loss: 1.5897 - val_accuracy: 0.4256 - 177s/epoch - 44s/step\n"
     ]
    }
   ],
   "source": [
    "#Now we have done the training is done, let's validate the learning with the test dataset\n",
    "history = model.fit(im_train, label_train, epochs=10, batch_size=4000, validation_data=(im_test, label_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 - 18s - loss: 1.5897 - accuracy: 0.4256 - 18s/epoch - 9s/step\n",
      "Testing Accuracy: 0.4256060719490051\n"
     ]
    }
   ],
   "source": [
    "#And finally let's see what is the testing accuracy\n",
    "\n",
    "score = model.score(im_test, label_test)\n",
    "print('Testing Accuracy:',score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEWCAYAAABsY4yMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3de5BcZ3nn8e/TPT33Hl1Go+mxZFkyxuoBg2V7cOyYUICBIAwGQmISMLXLbkWkioBJAcFmAxRbm122NssaQgIYMEsWrxPwZSHBEMVgcylARhICbI9k2ViXkXUZ6zr3S/ezf5wz0ow0Gs2lz5zu079PVdd0n+szXdLvvPOec95j7o6IiCRPKu4CREQkGgp4EZGEUsCLiCSUAl5EJKEU8CIiCaWAFxFJKAW8CGBm/9vM/sssl91jZq9Z6HZEoqaAFxFJKAW8iEhCKeClYoRdIx82s1+b2YCZfcXM2s3su2bWZ2YPm9myScvfbGZPmNkJM3vUzDonzbvKzLaH6/0TUH/Wvt5oZjvCdX9qZi+dZ81/amZPm9kxM/u2mV0UTjcz+19mdsTMToa/0xXhvDeY2ZNhbQfM7EPz+sKk6ingpdK8DXgtcDnwJuC7wEeBFQT/nt8PYGaXA/cCHwDagIeAfzazWjOrBf4f8H+A5cA3w+0Srns1cDfwHqAV+CLwbTOrm0uhZvZq4L8BtwAdwF7gH8PZrwNeEf4eS4G3A0fDeV8B3uPuWeAK4Adz2a/IBAW8VJq/dffD7n4A+DGwxd1/6e4jwIPAVeFybwe+4+7/5u5jwN8ADcDvAtcBGeBOdx9z9/uAX0zax58CX3T3Le5ecPevASPhenPxTuBud98e1ncHcL2ZrQXGgCyQB8zdu939YLjeGPAiM2tx9+Puvn2O+xUBFPBSeQ5Pej80zefm8P1FBC1mANy9COwHVoXzDvjUkfb2Tnp/CfDBsHvmhJmdAC4O15uLs2voJ2ilr3L3HwCfA/4OOGxmd5lZS7jo24A3AHvN7Idmdv0c9ysCKOAluZ4jCGog6PMmCOkDwEFgVThtwppJ7/cDf+3uSye9Gt393gXW0ETQ5XMAwN0/6+7XAC8m6Kr5cDj9F+7+ZmAlQVfSN+a4XxFAAS/J9Q3gJjO70cwywAcJull+CvwMGAfeb2Y1ZvYHwLWT1v0S8Gdm9jvhydAmM7vJzLJzrOH/Au82sw1h//1/JehS2mNmLwu3nwEGgGGgEJ4jeKeZLQm7lk4BhQV8D1LFFPCSSO6+C7gV+FvgeYITsm9y91F3HwX+APj3wHGC/voHJq27laAf/nPh/KfDZedaw/eBjwH3E/zV8ALgj8PZLQQHkuME3ThHCc4TALwL2GNmp4A/C38PkTkzPfBDRCSZ1IIXEUkoBbyISEJFGvBm9hfhnYSPm9m9ZlZ/4bVERKQUIgt4M1tFcFdhl7tfAaQ5c4JJREQiVrMI228wszGgkeC64PNasWKFr127NuKSRESSY9u2bc+7e9t08yILeHc/YGZ/A+wjuMNws7tvPns5M9sEbAJYs2YNW7dujaokEZHEMbO955sXZRfNMuDNwDqCW7abzOyc63nd/S5373L3rra2aQ9CIiIyD1GeZH0N8Ky794Z35D1AMNCTiIgsgigDfh9wnZk1hmN+3Ah0R7g/ERGZJMo++C1mdh+wnWDcj18Cd811O2NjY/T09DA8PFzqEstKfX09q1evJpPJxF2KiCREpFfRuPsngE8sZBs9PT1ks1nWrl3L1MH/ksPdOXr0KD09Paxbty7uckQkIcr+Ttbh4WFaW1sTG+4AZkZra2vi/0oRkcVV9gEPJDrcJ1TD7ygii6siAn4mRXeO9A3TNzwWdykiImWl4gPegN6+EU4ORRPwJ06c4O///u/nvN4b3vAGTpw4EUFFIiKzU/kBb0Z9Js3wWDGS7Z8v4AuFmR+y89BDD7F06dJIahIRmY2ox6JZFA2ZNMcGRnH3kvdl33777TzzzDNs2LCBTCZDc3MzHR0d7NixgyeffJK3vOUt7N+/n+HhYW677TY2bdoEwNq1a9m6dSv9/f1s3LiRl7/85fz0pz9l1apVfOtb36KhoaGkdYqInK2iAv6T//wETz536pzp48UiI2NFGmvTcw74F13Uwife9OLzzv/Upz7F448/zo4dO3j00Ue56aabePzxx09fznj33XezfPlyhoaGeNnLXsbb3vY2Wltbp2xj9+7d3HvvvXzpS1/illtu4f777+fWW/UUNhGJVkUF/PmkwlAvOqQjvhjl2muvnXKt+mc/+1kefPBBAPbv38/u3bvPCfh169axYcMGAK655hr27NkTbZEiIlRYwJ+vpV0oOk88d5L2lnraW6J9pkhTU9Pp948++igPP/wwP/vZz2hsbOSVr3zltNey19XVnX6fTqcZGhqKtEYREUjASVaAdMqoq0kzPDbzic/5yGaz9PX1TTvv5MmTLFu2jMbGRnbu3MnPf/7zku9fRGS+KqoFP5P6TCqSK2laW1u54YYbuOKKK2hoaKC9vf30vNe//vV84Qtf4KUvfSnr16/nuuuuK/n+RUTmy9w97hpO6+rq8rMf+NHd3U1nZ+cF1z18apjDp4Z58UVLSKcq867Q2f6uIiITzGybu3dNNy8RXTQA9Zk0ACPjpe+mERGpRAkK+OBXiaIfXkSkEiUm4GvTKVJmkd3RKiJSaRIT8BNDFgypBS8iAiQo4GHiSpoC5XTiWEQkLgkL+DSFojNeUMCLiCQq4BvCK2mGSnglzXyHCwa48847GRwcLFktIiJzkaiAr4vgShoFvIhUqsTcyQpQk0pRmy7tHa2Thwt+7Wtfy8qVK/nGN77ByMgIb33rW/nkJz/JwMAAt9xyCz09PRQKBT72sY9x+PBhnnvuOV71qlexYsUKHnnkkZLVJCIyG5UV8N+9HQ79ZsZF1o4VKOKQmeWvlnsJbPzUeWdPHi548+bN3HfffTz22GO4OzfffDM/+tGP6O3t5aKLLuI73/kOEIxRs2TJEj796U/zyCOPsGLFiln/iiIipZKoLhqAVAqKRXBKf6J18+bNbN68mauuuoqrr76anTt3snv3bl7ykpfw8MMP85GPfIQf//jHLFmypOT7FhGZq8ha8Ga2HvinSZMuBT7u7nfOe6MztLQnDA6Osu/YIC9cmaWhNj3vXU3H3bnjjjt4z3vec868bdu28dBDD3HHHXfwute9jo9//OMl3beIyFxF1oJ3913uvsHdNwDXAIPAg1Htb8LEmDTDJbqSZvJwwb//+7/P3XffTX9/PwAHDhzgyJEjPPfcczQ2NnLrrbfyoQ99iO3bt5+zrojIYlusPvgbgWfcfW/UO6qrSWFmJbuSZvJwwRs3buQd73gH119/PQDNzc18/etf5+mnn+bDH/4wqVSKTCbD5z//eQA2bdrExo0b6ejo0ElWEVl0izJcsJndDWx3989NM28TsAlgzZo11+zdO/UYMJ8hdHcf7qMmnWLdiqYLL1xGNFywiMxVrMMFm1ktcDPwzenmu/td7t7l7l1tbW0l2Wd9JpqnO4mIVJLFuIpmI0Hr/fAi7AsIAn6sUGS8oJElRaR6LUbA/wlw70I2MNdupEocG14DpIlIqUUa8GbWCLwWeGC+26ivr+fo0aNzCsDTV9JUyNjw7s7Ro0epr6+PuxQRSZBIr6Jx90GgdSHbWL16NT09PfT29s5pvedPDtF/OE1vY+1Cdr9o6uvrWb16ddxliEiClP1QBZlMhnXr1s15vb/+8hZODQ/y7T9/eQRViYiUv8QNVTAhn8uy61AfhaL6tkWkOiU34DtaGBkvsufoQNyliIjEIrkBn8sCsPOghgoQkeqU2IC/bGUzKYNdh07FXYqISCwSG/D1mTSXtjXTfUgteBGpTokNeAi6aXaqBS8iVSrRAd/Z0cL+Y0P0DY/FXYqIyKJLdMBPnGh96rC6aUSk+iQ74DtaAOjWlTQiUoUSHfAXLaknW1+jfngRqUqJDngzO31Hq4hItUl0wAPkcy3sPNin4XhFpOokP+A7svSNjHPgxFDcpYiILKrkB3wuONGqIQtEpNokPuDXT4xJoxOtIlJlEh/wzXU1rFneqCELRKTqJD7gIWjF7zyoFryIVJeqCPjOXJZnnx+oqIdwi4gsVFUEfL6jhaLD00f64y5FRGTRVEfAhydau9VNIyJVpCoC/pLWJuozKXbqRKuIVJGqCPh0yljfrrHhRaS6RBrwZrbUzO4zs51m1m1m10e5v5nkcy10a8gCEakiUbfgPwN8z93zwJVAd8T7O6/1uSzHBkbp7R+JqwQRkUUVWcCbWQvwCuArAO4+6u4notrfheQ7ghOtGllSRKpFlC34S4Fe4Ktm9ksz+7KZNUW4vxlpTBoRqTZRBnwNcDXweXe/ChgAbj97ITPbZGZbzWxrb29vZMUsb6qlvaWObp1oFZEqEWXA9wA97r4l/HwfQeBP4e53uXuXu3e1tbVFWM6ZseFFRKpBZAHv7oeA/Wa2Ppx0I/BkVPubjXxHlqeP9DNWKMZZhojIoqiJePvvA+4xs1rgt8C7I97fjDpzLYwWijz7/ACXt2fjLEVEJHKRBry77wC6otzHXKyfNGSBAl5Ekq4q7mSd8IK2ZmpSpkslRaQqVFXA19akuGxls8akEZGqUFUBD8HIknr4h4hUg+oL+I4Wnjs5zMnBsbhLERGJVPUFvB7CLSJVouoCvrMjHLJA/fAiknBVF/Ars3UsbcyoBS8iiVd1AW9mwYlWteBFJOGqLuAhGJNm16E+ikU9/ENEkqsqA76zI8vgaIH9xwfjLkVEJDJVGfATY8N3a2RJEUmwqgz4y9uzmOlSSRFJtqoM+IbaNOtamzQ2vIgkWlUGPAQjS6oFLyJJVrUBn8+1sPfYIIOj43GXIiISieoN+I4s7vDU4f64SxERiUTVBnxneCWNRpYUkaSq2oBfvayBptq07mgVkcSq2oBPpYz1uSzdasGLSEJVbcBDMDb8zkN9uGvIAhFJnuoO+FyWk0NjHDo1HHcpIiIlV+UBr7HhRSS5qjrg10883Ul3tIpIAlV1wC9pyLBqaYPuaBWRRKqJcuNmtgfoAwrAuLt3Rbm/+cjnsmrBi0giRRrwoVe5+/OLsJ95yXdk+eFTvYyMF6irScddjohIyVR1Fw0EJ1rHi84zRwbiLkVEpKSiDngHNpvZNjPbNN0CZrbJzLaa2dbe3t6IyzlXfuJEq/rhRSRhog74G9z9amAj8F4ze8XZC7j7Xe7e5e5dbW1tEZdzrnUrmqhNp9ilSyVFJGEiDXh3fy78eQR4ELg2yv3NR006xQvbm+lWwItIwkQW8GbWZGbZiffA64DHo9rfQuRzLRpVUkQSJ8oWfDvwEzP7FfAY8B13/16E+5u3zo4sR/pGONo/EncpIiIlM6uAN7PbzKzFAl8xs+1m9rqZ1nH337r7leHrxe7+16UpufQmhixQP7yIJMlsW/D/wd1PEXSztAHvBj4VWVWLLN8RXEmjfngRSZLZBryFP98AfNXdfzVpWsVb0VzHiuZa9cOLSKLMNuC3mdlmgoD/1/DkaTG6shZfPtfCrsNqwYtIcsw24P8jcDvwMncfBDIE3TSJkc9l2XWoj0JRD/8QkWSYbcBfD+xy9xNmdivwV8DJ6MpafPmOFkbGi+w5qiELRCQZZhvwnwcGzexK4C+BvcA/RFZVDPIaG15EEma2AT/uwYNL3wx8xt0/A2SjK2vxXbaymXTKNCaNiCTGbIcL7jOzO4B3Ab9nZmmCfvjEqM+kuXRFE91qwYtIQsy2Bf92YITgevhDwCrgf0RWVUzW57JqwYtIYswq4MNQvwdYYmZvBIbdPVF98ACdHS30HB+ib3gs7lJERBZstkMV3EIwnswfAbcAW8zsD6MsLA4TJ1qf0vXwIpIAs+2D/08E18AfATCzNuBh4L6oCotDviMYk6b7YB/XXLI85mpERBZmtn3wqYlwDx2dw7oV46Il9WTra9QPLyKJMNsW/PfM7F+Be8PPbwceiqak+JgZnbkWXQsvIokwq4B39w+b2duAGwgGGbvL3R+MtLKY5DuyPLD9AO6OWWLGUxORKjTbFjzufj9wf4S1lIV8roX+kb30HB/i4uWNcZcjIjJvMwa8mfUB042+ZYC7e0skVcVofXglza5DfQp4EaloM54odfesu7dM88omMdzhTMDrRKuIVLrEXQmzUM11NaxZ3qinO4lIxVPATyOfy+rpTiJS8RTw08h3tPDs8wMMjxXiLkVEZN4U8NPozGUpOuw+3B93KSIi86aAn8bpIQt0olVEKljkAW9maTP7pZn9S9T7KpU1yxupz6TYpROtIlLBFqMFfxvQvQj7KZl0yljfrrHhRaSyRRrwZrYauAn4cpT7iUI+10L3wT6CJxWKiFSeqFvwdxI8pLt4vgXMbJOZbTWzrb29vRGXM3v5jizHBkbp7R+JuxQRkXmJLODDJz8dcfdtMy3n7ne5e5e7d7W1tUVVzpzlc8GJVo0sKSKVKsoW/A3AzWa2B/hH4NVm9vUI91dSeQ1ZICIVLrKAd/c73H21u68F/hj4gbvfGtX+Sm1ZUy25lnq14EWkYuk6+Bmsz2XZqUslRaRCLUrAu/uj7v7GxdhXKeU7sjx9pJ+xwnnPEYuIlC214GfQmWthtFDk2ecH4i5FRGTOFPAzyHcEJ1q7NbKkiFQgBfwMLl3RTCZt6ocXkYqkgJ9BbU2KF7Q1a2x4EalICvgL6OxoUQteRCqSAv4C1ueyHDw5zMnBsbhLERGZEwX8BeiOVhGpVAr4C+gMH/6hbhoRqTQK+AtYma1jWWNGLXgRqTgK+Asws9Njw4uIVBIF/CzkO7LsOtRHsaiHf4hI5VDAz0JnroWhsQL7jg3GXYqIyKwp4Gdh/ekradRNIyKVQwE/C5e3ZzHTpZIiUlkU8LPQUJtmXWuTHv4hIhVFAT9L+Y6sWvAiUlEU8LOUz7Ww99ggAyPjcZciIjIrCvhZyueyuMNTh9VNIyKVQQE/SxqyQEQqjQJ+llYtbaCpNs0uBbyIVAgF/CylUsb6XFaP7xORiqGAn4N8+PAPdw1ZICLlTwE/B525LCeHxjh0ajjuUkRELiiygDezejN7zMx+ZWZPmNkno9rXYslPnGjVDU8iUgGibMGPAK929yuBDcDrzey6CPcXuYkxabp1w5OIVICaqDbsQUd1f/gxE74quvO6pT7DqqUNasGLSEWItA/ezNJmtgM4Avybu2+ZZplNZrbVzLb29vZGWU5J5HNZXSopIhUh0oB394K7bwBWA9ea2RXTLHOXu3e5e1dbW1uU5ZREviPLM739jIwX4i5FRGRGi3IVjbufAB4FXr8Y+4tSPtfCeNF55shA3KWIiMwoyqto2sxsafi+AXgNsDOq/S2Wzo6Jh3/oRKuIlLfITrICHcDXzCxNcCD5hrv/S4T7WxRrW5uorUlpTBoRKXtRXkXza+CqqLYfl5p0isvbmzVkgYiUPd3JOg/5XIta8CJS9hTw85DPZentG+Fo/0jcpYiInJcCfh7yuWDIAl0PLyLlTAE/D/mOiSELFPAiUr4U8POwormOFc117NSJVhEpYwr4eersyOpEq4iUNQX8POVzWZ463Md4oRh3KSIi01LAz1M+18LIeJE9RwfjLkVEZFoK+HmaGBteV9KISLlSwM/TZSubSadMY9KISNlSwM9TfSbNpSua6NbDP0SkTCngFyDf0aIWvIiULQX8AuRzWXqOD3FqeCzuUkREzqGAX4CJseGf0olWESlDCvgFmBiTRkMWiEg5UsAvQMeSerL1NexSP7yIlCEF/AKYGZ25FnbqShoRKUMK+AXKh2PSuHvcpYiITKGAX6B8roX+kXF6jg/FXYqIyBQK+AWaGBteI0uKSLlRwC/Q+vYw4DU2vIiUGQX8AjXV1XBJa6Na8CJSdhTwJbC+PashC0Sk7EQW8GZ2sZk9YmbdZvaEmd0W1b7ilu9o4dnnBxgeK8RdiojIaVG24MeBD7p7J3Ad8F4ze1GE+4tNZy5L0WH34f64SxEROS2ygHf3g+6+PXzfB3QDq6LaX5zyHRNDFqibRkTKx6L0wZvZWuAqYMs08zaZ2VYz29rb27sY5ZTcmuWNNGTSuqNVRMpK5AFvZs3A/cAH3P2cJq673+XuXe7e1dbWFnU5kUinjMtzOtEqIuUl0oA3swxBuN/j7g9Eua+4deY0ZIGIlJcor6Ix4CtAt7t/Oqr9APDof4ctX4Rd34MjO2F0MNLdTWd9LsuxgVF6+0cWfd8iItOpiXDbNwDvAn5jZjvCaR9194dKupdiAX72ORg5q3ukaSUsWwvLLgl+Lr3kzOeWVZBKl7SMibHhP/GtJ7jy4qWsWd7ImuWNXLy8kSUNmZLuS0RkNiILeHf/CWBRbf+0VBpu3wcDvXB8LxzfAyf2BD+P74V9W+Dx+8GLk9bJwJLV5zkArIWGZWBzK/3Ki5fwuy9oZcuzx/ju44emzFvSkJkS+Bcvbzj9+aKlDWTSut9MRErPyqnPuKury7du3Vr6DRfG4OT+SQeAvWcOACf2wuDRqcvXtYSBPyn0Jw4AS9dApn7G3fUNj7H/2BD7jg2y/9gg+8LX/mOD9BwfYrRw5mCTMrhoacOUA8CaSa+ljRlsjgcbEakeZrbN3bumnVcVAX8hI33Th//E5/HhqctnO6Z2+Zw+AFwCzTlIn/8Po2LROdw3zL6jg+ccAPYdG+L5s/rws3U1Z0K/deoBYNXSBmpr1PoXqWYK+IUoFmHgyPkPAKcOAJO/Q4OmFUHQN6+EbA6a24NXtj2Yng0/1zads7vB0fHTrf/p/gIYGT/T+jeDjpb6qa3+1kZWL2tgRXMdy5tqaa6r0V8AIgk2U8BHeZI1GVKpIKSzOVjzO+fOHx+Bkz1nwr/vUPDqPwL9h+BId3CAKI6fu25t9kzYN7dDNkdjczvrm9tZn22HF+aCA0LjcjCjWHSe7x+Z1OI/E/w/2t3L4VPnXsFTm06xvKmW5U21tDbXnnnfVMvyprop01ubammpz5BK6YAgkgQK+IWqqYPWFwSv8ykWYehYGPyHg9eU94fh4A7YfQRGpxnPJpWB5nZS2XZWNgevrmwOlq6E1RN/EVzCcF0rPafG2H98iGP9oxwbGOXowCjHBkZOv993bJBj/aP0jUxzwCG4aWtZ48QBoJblzWfeT3dQWNZYS1oHBJGypIBfDKlU0G3TtAK4YuZlR/rPfxDoPxR0De1/DAafP2fVeuCyxlYua1oZnAiuqQ8OQDX10FAHzXWnP4+naxkqZhgs1DBQSNNfSHNqPM3JsRQnRtOcGDGODRpHj6d4ZggeG0kxQoZRr2GEWkbIMEKGMathacPEASAI/2XhwWBpY4a6mhS1E690etL7FHWZ8OeUZYKfdTVpMmlT95LIAijgy01dc/Ca6S8CCK4M6j9y1sEg7BYa6A26jsaHYWwIho6Hn0dOT68ZHyE7PkR28uWjM9Z1/lljXsvYQIbRgVqGPcOw1zBUrGGUGoqkKJCiiFEkRdGNAinGMEbOnkcwz8OfRVJgKcxSkEpjlsJS6WBaOo1ZGkulsFQNqVQwL5VKk0qnSaVSpFJpLB1Ms1SwnWAbwSs16b3ZxDqGhctPzE+FL7M06XQwPz2xzOlXsH46nSJlaVLpYNrkzxPHKsMwgnMo5z98TXNu7Lzny84zfdrly+ScW9mc+5tUx5SafNq353x/511njtPTdfDC11yo2DlTwFeqdAaWrApeC1EYh8KZ4A9eE+9Hz/oc/pyy/AiZ8WEy4yM0TjqA+Pgw46MjeLFA0Yt4oYAXw5cXw/dF3EehWMS9EHRleRFOvy+AF7HwBeH7YhErBO9TFDF3UhQwPDwslEt4iMzO8dQyln18T8m3q4Cvduma4DXNFT0LYUBs9++6h6/gAOHFcQqFIuOFAsVikfFikeJ4gUIxfBWcQqFAsVigWCgE84sFioUihWKwTqEQTCsUCnjRw+nhMuEBrFgsBsuGB6dgneDlxXHc7fShx/GgxNOfpzbuJuZNXgagOGkFn1gPx93C9c4sfXq+T96+z/kmvqg4duZ39okfwRv3KZMnreNnLR/+POsvgrPXP+eQP2V/NmnypO/Gzp1u022LM12Jk9f3KX+hTd3WxKYnlm+oq+V952x34RTwkjxm4X/O4B4Bo44a9I9dqo/ukhERSSgFvIhIQingRUQSSgEvIpJQCngRkYRSwIuIJJQCXkQkoRTwIiIJVVbjwZtZL7B3nquvAM4dgas66buYSt/HVPo+zkjCd3GJu7dNN6OsAn4hzGzr+Qa9rzb6LqbS9zGVvo8zkv5dqItGRCShFPAiIgmVpIC/K+4Cyoi+i6n0fUyl7+OMRH8XiemDFxGRqZLUghcRkUkU8CIiCVXxAW9mrzezXWb2tJndHnc9cTKzi83sETPrNrMnzOy2uGuKm5mlzeyXZvYvcdcSNzNbamb3mdnO8N/I9XHXFCcz+4vw/8njZnavmdXHXVOpVXTAm1ka+DtgI/Ai4E/M7EXxVhWrceCD7t4JXAe8t8q/D4DbgO64iygTnwG+5+554Eqq+Hsxs1XA+4Eud78CSAN/HG9VpVfRAQ9cCzzt7r9191HgH4E3x1xTbNz9oLtvD9/3EfwHXuBTuSuXma0GbgK+HHctcTOzFuAVwFcA3H3U3U/EW1XsaoAGM6sBGoHnYq6n5Co94FcB+yd97qGKA20yM1sLXAVsibeSWN0J/CVQjLuQMnAp0At8Neyy+rKZlfZJ6xXE3Q8AfwPsAw4CJ919c7xVlV6lB/x0j4ev+us+zawZuB/4gLufirueOJjZG4Ej7r4t7lrKRA1wNfB5d78KGACq9pyVmS0j+Gt/HXAR0GRmt8ZbVelVesD3ABdP+ryaBP6ZNRdmliEI93vc/YG464nRDcDNZraHoOvu1Wb29XhLilUP0OPuE3/R3UcQ+NXqNcCz7t7r7mPAA8DvxlxTyVV6wP8CeKGZrTOzWoKTJN+OuabYmJkR9LF2u/un464nTu5+h7uvdve1BP8ufuDuiWuhzZa7HwL2m9n6cNKNwJMxlhS3fcB1ZtYY/r+5kQSedK6Ju4CFcPdxM/tz4F8JzoLf7e5PxFxWnG4A3gX8xsx2hNM+6u4PxaY+wO4AAAHzSURBVFiTlI/3AfeEjaHfAu+OuZ7YuPsWM7sP2E5w9dkvSeCwBRqqQEQkoSq9i0ZERM5DAS8iklAKeBGRhFLAi4gklAJeRCShFPAiJWBmr9SIlVJuFPAiIgmlgJeqYma3mtljZrbDzL4Yjhffb2b/08y2m9n3zawtXHaDmf3czH5tZg+G45dgZpeZ2cNm9qtwnReEm2+eNN76PeEdkiKxUcBL1TCzTuDtwA3uvgEoAO8EmoDt7n418EPgE+Eq/wB8xN1fCvxm0vR7gL9z9ysJxi85GE6/CvgAwbMJLiW4s1gkNhU9VIHIHN0IXAP8ImxcNwBHCIYT/qdwma8DD5jZEmCpu/8wnP414JtmlgVWufuDAO4+DBBu7zF37wk/7wDWAj+J/tcSmZ4CXqqJAV9z9zumTDT72FnLzTR+x0zdLiOT3hfQ/y+JmbpopJp8H/hDM1sJYGbLzewSgv8Hfxgu8w7gJ+5+EjhuZr8XTn8X8MNwfP0eM3tLuI06M2tc1N9CZJbUwpCq4e5PmtlfAZvNLAWMAe8lePjFi81sG3CSoJ8e4N8BXwgDfPLoi+8Cvmhm/zncxh8t4q8hMmsaTVKqnpn1u3tz3HWIlJq6aEREEkoteBGRhFILXkQkoRTwIiIJpYAXEUkoBbyISEIp4EVEEur/A4hM1eKK3vtcAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Let's now plot the differnces in the losses and the accuracies between the training and the test\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}