CIFAR - Traditional Filters.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Load the prepared dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['cifar10', 'mnist']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "DATA_DIR = 'C:/Users/MINH LE/Desktop/TraditionalvsCNN/dataset/'\n",
    "\n",
    "list_files = os.listdir(DATA_DIR)\n",
    "number_files = len(list_files)\n",
    "list_files"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.1. CIFAR10 Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**CLTP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import glob, cv2\n",
    "cltp = []\n",
    "\n",
    "CLTP_DIR = DATA_DIR + 'cifar10/cltp/train/'\n",
    "for img in glob.glob(CLTP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    cltp.append(n)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cltp = np.asarray(cltp, dtype=np.float32)\n",
    "cltp.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**CSLBP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cslbp = []\n",
    "\n",
    "CSLBP_DIR = DATA_DIR + 'cifar10/cslbp/train/'\n",
    "for img in glob.glob(CSLBP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    cslbp.append(n)\n",
    "    \n",
    "cslbp = np.asarray(cslbp, dtype=np.float32)\n",
    "cslbp.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**GAUSSIAN FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gauss_filter = []\n",
    "\n",
    "GAUSSIAN_DIR = DATA_DIR + 'cifar10/gaussian/train/'\n",
    "for img in glob.glob(GAUSSIAN_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    gauss_filter.append(n)\n",
    "    \n",
    "gauss_filter = np.asarray(gauss_filter, dtype=np.float32)\n",
    "gauss_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**GLTP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gltp_filter = []\n",
    "\n",
    "GLTP_DIR = DATA_DIR + 'cifar10/gltp/train/'\n",
    "for img in glob.glob(GLTP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    gltp_filter.append(n)\n",
    "    \n",
    "gltp_filter = np.asarray(gltp_filter, dtype=np.float32)\n",
    "gltp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**LAPLACIAN FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "laplacian_filter = []\n",
    "\n",
    "LAPLACIAN_DIR = DATA_DIR + 'cifar10/laplacian/train/'\n",
    "for img in glob.glob(LAPLACIAN_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    laplacian_filter.append(n)\n",
    "    \n",
    "laplacian_filter = np.asarray(laplacian_filter, dtype=np.float32)\n",
    "laplacian_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**LOCAL BINARY PATTERN FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lbp_filter = []\n",
    "\n",
    "LBP_DIR = DATA_DIR + 'cifar10/lbp/train/'\n",
    "for img in glob.glob(LBP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    lbp_filter.append(n)\n",
    "    \n",
    "lbp_filter = np.asarray(lbp_filter, dtype=np.float32)\n",
    "lbp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**LOCAL TERNARY PATTERN**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ltp_filter = []\n",
    "\n",
    "LTP_DIR = DATA_DIR + 'cifar10/ltp/train/'\n",
    "for img in glob.glob(LTP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    ltp_filter.append(n)\n",
    "    \n",
    "ltp_filter = np.asarray(ltp_filter, dtype=np.float32)\n",
    "ltp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**MC-LBP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mclbp_filter = []\n",
    "\n",
    "MCLBP_DIR = DATA_DIR + 'cifar10/mclbp/train/'\n",
    "for img in glob.glob(MCLBP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    mclbp_filter.append(n)\n",
    "    \n",
    "mclbp_filter = np.asarray(mclbp_filter, dtype=np.float32)\n",
    "mclbp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**PREWITT X FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prewittx_filter = []\n",
    "\n",
    "PREWITTX_DIR = DATA_DIR + 'cifar10/prewittx/train/'\n",
    "for img in glob.glob(PREWITTX_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    prewittx_filter.append(n)\n",
    "    \n",
    "prewittx_filter = np.asarray(prewittx_filter, dtype=np.float32)\n",
    "prewittx_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**PREWITT Y FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prewitty_filter = []\n",
    "\n",
    "PREWITTY_DIR = DATA_DIR + 'cifar10/prewitty/train/'\n",
    "for img in glob.glob(PREWITTY_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    prewitty_filter.append(n)\n",
    "    \n",
    "prewitty_filter = np.asarray(prewitty_filter, dtype=np.float32)\n",
    "prewitty_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SCBP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scbp_filter = []\n",
    "\n",
    "SCBP_DIR = DATA_DIR + 'cifar10/scbp/train/'\n",
    "for img in glob.glob(SCBP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    scbp_filter.append(n)\n",
    "    \n",
    "scbp_filter = np.asarray(scbp_filter, dtype=np.float32)\n",
    "scbp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SCS-LTP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scsltp_filter = []\n",
    "\n",
    "SCSLTP_DIR = DATA_DIR + 'cifar10/scsltp/train/'\n",
    "for img in glob.glob(SCSLTP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    scsltp_filter.append(n)\n",
    "    \n",
    "scsltp_filter = np.asarray(scsltp_filter, dtype=np.float32)\n",
    "scsltp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SILTP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "siltp_filter = []\n",
    "\n",
    "SILTP_DIR = DATA_DIR + 'cifar10/siltp/train/'\n",
    "for img in glob.glob(SILTP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    siltp_filter.append(n)\n",
    "    \n",
    "siltp_filter = np.asarray(siltp_filter, dtype=np.float32)\n",
    "siltp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SOBEL X FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sobelx_filter = []\n",
    "\n",
    "SOBELX_DIR = DATA_DIR + 'cifar10/sobelx/train/'\n",
    "for img in glob.glob(SOBELX_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    sobelx_filter.append(n)\n",
    "    \n",
    "sobelx_filter = np.asarray(sobelx_filter, dtype=np.float32)\n",
    "sobelx_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**SOBEL Y FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sobely_filter = []\n",
    "\n",
    "SOBELY_DIR = DATA_DIR + 'cifar10/sobely/train/'\n",
    "for img in glob.glob(SOBELY_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    sobely_filter.append(n)\n",
    "    \n",
    "sobely_filter = np.asarray(sobely_filter, dtype=np.float32)\n",
    "sobely_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**XCSLTP FILTER**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 30, 30, 3)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xcsltp_filter = []\n",
    "\n",
    "XCSLTP_DIR = DATA_DIR + 'cifar10/xcsltp/train/'\n",
    "for img in glob.glob(XCSLTP_DIR + '*.jpg'):\n",
    "    n = cv2.imread(img)\n",
    "    xcsltp_filter.append(n)\n",
    "    \n",
    "xcsltp_filter = np.asarray(xcsltp_filter, dtype=np.float32)\n",
    "xcsltp_filter.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Training Phase"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import Flatten\n",
    "from tensorflow.keras.layers import MaxPooling2D\n",
    "from tensorflow.keras.layers import Input\n",
    "from tensorflow.keras.layers import Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([50000, 30, 30, 48])"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data = tf.keras.layers.Concatenate()([cltp, \n",
    "                                            cslbp, \n",
    "                                            gauss_filter, \n",
    "                                            gltp_filter,\n",
    "                                            laplacian_filter,\n",
    "                                            lbp_filter,\n",
    "                                            ltp_filter,\n",
    "                                            mclbp_filter,\n",
    "                                            prewittx_filter,\n",
    "                                            prewitty_filter,\n",
    "                                            scsltp_filter,\n",
    "                                            siltp_filter,\n",
    "                                            sobelx_filter,\n",
    "                                            sobely_filter,\n",
    "                                            xcsltp_filter,\n",
    "                                            scbp_filter])\n",
    "train_data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "cifar10 = keras.datasets.cifar10\n",
    "(_, train_labels), (test_images, test_labels) = cifar10.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: keras in c:\\users\\minh le\\anaconda3\\lib\\site-packages (2.4.3)\n",
      "Requirement already satisfied: scipy>=0.14 in c:\\users\\minh le\\anaconda3\\lib\\site-packages (from keras) (1.6.0)\n",
      "Requirement already satisfied: numpy>=1.9.1 in c:\\users\\minh le\\anaconda3\\lib\\site-packages (from keras) (1.19.5)\n",
      "Requirement already satisfied: pyyaml in c:\\users\\minh le\\anaconda3\\lib\\site-packages (from keras) (5.3.1)\n",
      "Requirement already satisfied: h5py in c:\\users\\minh le\\anaconda3\\lib\\site-packages (from keras) (2.10.0)\n",
      "Requirement already satisfied: six in c:\\users\\minh le\\anaconda3\\lib\\site-packages (from h5py->keras) (1.15.0)\n"
     ]
    }
   ],
   "source": [
    "!pip install keras"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Experiment 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "49/49 [==============================] - 25s 464ms/step - loss: 4986.8691 - accuracy: 0.1022\n",
      "Epoch 2/300\n",
      "49/49 [==============================] - 21s 429ms/step - loss: 24.7451 - accuracy: 0.1009\n",
      "Epoch 3/300\n",
      "49/49 [==============================] - 21s 424ms/step - loss: 2.3026 - accuracy: 0.1000\n",
      "Epoch 4/300\n",
      "49/49 [==============================] - 19s 395ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 5/300\n",
      "49/49 [==============================] - 20s 403ms/step - loss: 2.3026 - accuracy: 0.0986\n",
      "Epoch 6/300\n",
      "49/49 [==============================] - 20s 413ms/step - loss: 2.3026 - accuracy: 0.1012\n",
      "Epoch 7/300\n",
      "49/49 [==============================] - 21s 426ms/step - loss: 2.3026 - accuracy: 0.0997\n",
      "Epoch 8/300\n",
      "49/49 [==============================] - 20s 414ms/step - loss: 2.3026 - accuracy: 0.1001\n",
      "Epoch 9/300\n",
      "49/49 [==============================] - 22s 444ms/step - loss: 2.3026 - accuracy: 0.0997\n",
      "Epoch 10/300\n",
      "49/49 [==============================] - 20s 410ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 11/300\n",
      "49/49 [==============================] - 19s 395ms/step - loss: 2.3026 - accuracy: 0.0984\n",
      "Epoch 12/300\n",
      "49/49 [==============================] - 21s 438ms/step - loss: 2.3025 - accuracy: 0.1011\n",
      "Epoch 13/300\n",
      "49/49 [==============================] - 21s 439ms/step - loss: 2.3026 - accuracy: 0.1006\n",
      "Epoch 14/300\n",
      "49/49 [==============================] - 21s 419ms/step - loss: 2.3025 - accuracy: 0.1000\n",
      "Epoch 15/300\n",
      "49/49 [==============================] - 19s 393ms/step - loss: 2.3025 - accuracy: 0.1002\n",
      "Epoch 16/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0972\n",
      "Epoch 17/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3025 - accuracy: 0.1000\n",
      "Epoch 18/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 19/300\n",
      "49/49 [==============================] - 18s 359ms/step - loss: 2.3025 - accuracy: 0.1003\n",
      "Epoch 20/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3025 - accuracy: 0.0990\n",
      "Epoch 21/300\n",
      "49/49 [==============================] - 18s 363ms/step - loss: 2.3026 - accuracy: 0.0986\n",
      "Epoch 22/300\n",
      "49/49 [==============================] - 18s 373ms/step - loss: 2.3026 - accuracy: 0.1000\n",
      "Epoch 23/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 24/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.0975\n",
      "Epoch 25/300\n",
      "49/49 [==============================] - 19s 389ms/step - loss: 2.3025 - accuracy: 0.0980\n",
      "Epoch 26/300\n",
      "49/49 [==============================] - 19s 397ms/step - loss: 2.3024 - accuracy: 0.1015\n",
      "Epoch 27/300\n",
      "49/49 [==============================] - 20s 401ms/step - loss: 2.3025 - accuracy: 0.1000\n",
      "Epoch 28/300\n",
      "49/49 [==============================] - 19s 394ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 29/300\n",
      "49/49 [==============================] - 18s 378ms/step - loss: 2.3026 - accuracy: 0.0968\n",
      "Epoch 30/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0993\n",
      "Epoch 31/300\n",
      "49/49 [==============================] - 18s 358ms/step - loss: 2.3025 - accuracy: 0.1000\n",
      "Epoch 32/300\n",
      "49/49 [==============================] - 18s 359ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 33/300\n",
      "49/49 [==============================] - 18s 362ms/step - loss: 2.3025 - accuracy: 0.0996\n",
      "Epoch 34/300\n",
      "49/49 [==============================] - 18s 364ms/step - loss: 2.3026 - accuracy: 0.0969\n",
      "Epoch 35/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.0981\n",
      "Epoch 36/300\n",
      "49/49 [==============================] - 18s 373ms/step - loss: 2.3026 - accuracy: 0.0985\n",
      "Epoch 37/300\n",
      "49/49 [==============================] - 19s 383ms/step - loss: 2.3026 - accuracy: 0.0993\n",
      "Epoch 38/300\n",
      "49/49 [==============================] - 19s 384ms/step - loss: 2.3025 - accuracy: 0.0972\n",
      "Epoch 39/300\n",
      "49/49 [==============================] - 19s 383ms/step - loss: 2.3025 - accuracy: 0.0968\n",
      "Epoch 40/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3025 - accuracy: 0.1005\n",
      "Epoch 41/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 42/300\n",
      "49/49 [==============================] - 18s 376ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 43/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3025 - accuracy: 0.1001\n",
      "Epoch 44/300\n",
      "49/49 [==============================] - 17s 355ms/step - loss: 2.3025 - accuracy: 0.0963\n",
      "Epoch 45/300\n",
      "49/49 [==============================] - 20s 402ms/step - loss: 2.3025 - accuracy: 0.0984\n",
      "Epoch 46/300\n",
      "49/49 [==============================] - 18s 378ms/step - loss: 2.3025 - accuracy: 0.1020\n",
      "Epoch 47/300\n",
      "49/49 [==============================] - 18s 373ms/step - loss: 2.3025 - accuracy: 0.0967\n",
      "Epoch 48/300\n",
      "49/49 [==============================] - 18s 368ms/step - loss: 2.3026 - accuracy: 0.0955\n",
      "Epoch 49/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3025 - accuracy: 0.0983\n",
      "Epoch 50/300\n",
      "49/49 [==============================] - 19s 382ms/step - loss: 2.3026 - accuracy: 0.0970\n",
      "Epoch 51/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 52/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0985\n",
      "Epoch 53/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 54/300\n",
      "49/49 [==============================] - 19s 389ms/step - loss: 2.3026 - accuracy: 0.0971\n",
      "Epoch 55/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.1012\n",
      "Epoch 56/300\n",
      "49/49 [==============================] - 20s 398ms/step - loss: 2.3025 - accuracy: 0.1017\n",
      "Epoch 57/300\n",
      "49/49 [==============================] - 19s 383ms/step - loss: 2.3024 - accuracy: 0.0970\n",
      "Epoch 58/300\n",
      "49/49 [==============================] - 19s 378ms/step - loss: 2.3026 - accuracy: 0.0990\n",
      "Epoch 59/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0979\n",
      "Epoch 60/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 61/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.0975\n",
      "Epoch 62/300\n",
      "49/49 [==============================] - 18s 364ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 63/300\n",
      "49/49 [==============================] - 18s 367ms/step - loss: 2.3025 - accuracy: 0.0987\n",
      "Epoch 64/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.1036\n",
      "Epoch 65/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 66/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0965\n",
      "Epoch 67/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.1006\n",
      "Epoch 68/300\n",
      "49/49 [==============================] - 19s 382ms/step - loss: 2.3025 - accuracy: 0.1002\n",
      "Epoch 69/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3025 - accuracy: 0.0985\n",
      "Epoch 70/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0997\n",
      "Epoch 71/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.1023\n",
      "Epoch 72/300\n",
      "49/49 [==============================] - 19s 382ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 73/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3025 - accuracy: 0.0983\n",
      "Epoch 74/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0984\n",
      "Epoch 75/300\n",
      "49/49 [==============================] - 18s 359ms/step - loss: 2.3025 - accuracy: 0.1010\n",
      "Epoch 76/300\n",
      "49/49 [==============================] - 17s 349ms/step - loss: 2.3025 - accuracy: 0.0981\n",
      "Epoch 77/300\n",
      "49/49 [==============================] - 17s 357ms/step - loss: 2.3026 - accuracy: 0.0953\n",
      "Epoch 78/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3025 - accuracy: 0.1009\n",
      "Epoch 79/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.0963\n",
      "Epoch 80/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3025 - accuracy: 0.0979\n",
      "Epoch 81/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0991\n",
      "Epoch 82/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0967\n",
      "Epoch 83/300\n",
      "49/49 [==============================] - 19s 382ms/step - loss: 2.3025 - accuracy: 0.1004\n",
      "Epoch 84/300\n",
      "49/49 [==============================] - 19s 383ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 85/300\n",
      "49/49 [==============================] - 19s 385ms/step - loss: 2.3025 - accuracy: 0.1004\n",
      "Epoch 86/300\n",
      "49/49 [==============================] - 19s 388ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 87/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3025 - accuracy: 0.0978\n",
      "Epoch 88/300\n",
      "49/49 [==============================] - 18s 362ms/step - loss: 2.3026 - accuracy: 0.1000\n",
      "Epoch 89/300\n",
      "49/49 [==============================] - 17s 355ms/step - loss: 2.3027 - accuracy: 0.0998\n",
      "Epoch 90/300\n",
      "49/49 [==============================] - 17s 353ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 91/300\n",
      "49/49 [==============================] - 18s 359ms/step - loss: 2.3026 - accuracy: 0.0966\n",
      "Epoch 92/300\n",
      "49/49 [==============================] - 18s 361ms/step - loss: 2.3026 - accuracy: 0.1007\n",
      "Epoch 93/300\n",
      "49/49 [==============================] - 18s 361ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 94/300\n",
      "49/49 [==============================] - 18s 376ms/step - loss: 2.3026 - accuracy: 0.1023\n",
      "Epoch 95/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.0966\n",
      "Epoch 96/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 97/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0970\n",
      "Epoch 98/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 99/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 100/300\n",
      "49/49 [==============================] - 19s 397ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 101/300\n",
      "49/49 [==============================] - 20s 415ms/step - loss: 2.3026 - accuracy: 0.1027\n",
      "Epoch 102/300\n",
      "49/49 [==============================] - 19s 392ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 103/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 104/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 105/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.0969\n",
      "Epoch 106/300\n",
      "49/49 [==============================] - 17s 353ms/step - loss: 2.3026 - accuracy: 0.0978\n",
      "Epoch 107/300\n",
      "49/49 [==============================] - 18s 369ms/step - loss: 2.3026 - accuracy: 0.1000\n",
      "Epoch 108/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0967\n",
      "Epoch 109/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0978\n",
      "Epoch 110/300\n",
      "49/49 [==============================] - 18s 361ms/step - loss: 2.3026 - accuracy: 0.0982\n",
      "Epoch 111/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.1028\n",
      "Epoch 112/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 113/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.0965\n",
      "Epoch 114/300\n",
      "49/49 [==============================] - 19s 382ms/step - loss: 2.3026 - accuracy: 0.0945\n",
      "Epoch 115/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0993\n",
      "Epoch 116/300\n",
      "49/49 [==============================] - 19s 387ms/step - loss: 2.3026 - accuracy: 0.0971\n",
      "Epoch 117/300\n",
      "49/49 [==============================] - 19s 391ms/step - loss: 2.3026 - accuracy: 0.0968\n",
      "Epoch 118/300\n",
      "49/49 [==============================] - 19s 398ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 119/300\n",
      "49/49 [==============================] - 19s 384ms/step - loss: 2.3026 - accuracy: 0.0991\n",
      "Epoch 120/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0968\n",
      "Epoch 121/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0969\n",
      "Epoch 122/300\n",
      "49/49 [==============================] - 18s 363ms/step - loss: 2.3026 - accuracy: 0.1006\n",
      "Epoch 123/300\n",
      "49/49 [==============================] - 18s 364ms/step - loss: 2.3026 - accuracy: 0.1011\n",
      "Epoch 124/300\n",
      "49/49 [==============================] - 18s 364ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 125/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.1001\n",
      "Epoch 126/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.1003\n",
      "Epoch 127/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.1001\n",
      "Epoch 128/300\n",
      "49/49 [==============================] - 23s 465ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 129/300\n",
      "49/49 [==============================] - 22s 438ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 130/300\n",
      "49/49 [==============================] - 22s 447ms/step - loss: 2.3026 - accuracy: 0.0966\n",
      "Epoch 131/300\n",
      "49/49 [==============================] - 23s 474ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 132/300\n",
      "49/49 [==============================] - 23s 461ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 133/300\n",
      "49/49 [==============================] - 20s 415ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 134/300\n",
      "49/49 [==============================] - 20s 400ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 135/300\n",
      "49/49 [==============================] - 20s 415ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 136/300\n",
      "49/49 [==============================] - 21s 428ms/step - loss: 2.3026 - accuracy: 0.1018\n",
      "Epoch 137/300\n",
      "49/49 [==============================] - 19s 387ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 138/300\n",
      "49/49 [==============================] - 19s 386ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 139/300\n",
      "49/49 [==============================] - 19s 365ms/step - loss: 2.3026 - accuracy: 0.0989\n",
      "Epoch 140/300\n",
      "49/49 [==============================] - 19s 389ms/step - loss: 2.3026 - accuracy: 0.1022\n",
      "Epoch 141/300\n",
      "49/49 [==============================] - 21s 432ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 142/300\n",
      "49/49 [==============================] - 20s 404ms/step - loss: 2.3026 - accuracy: 0.0990\n",
      "Epoch 143/300\n",
      "49/49 [==============================] - 21s 421ms/step - loss: 2.3026 - accuracy: 0.0980\n",
      "Epoch 144/300\n",
      "49/49 [==============================] - 21s 428ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 145/300\n",
      "49/49 [==============================] - 22s 443ms/step - loss: 2.3026 - accuracy: 0.0986\n",
      "Epoch 146/300\n",
      "49/49 [==============================] - 21s 428ms/step - loss: 2.3026 - accuracy: 0.0984\n",
      "Epoch 147/300\n",
      "49/49 [==============================] - 20s 412ms/step - loss: 2.3026 - accuracy: 0.0972\n",
      "Epoch 148/300\n",
      "49/49 [==============================] - 19s 392ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 149/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.1004\n",
      "Epoch 150/300\n",
      "49/49 [==============================] - 19s 381ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 151/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.1006\n",
      "Epoch 152/300\n",
      "49/49 [==============================] - 18s 359ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 153/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 154/300\n",
      "49/49 [==============================] - 19s 388ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 155/300\n",
      "49/49 [==============================] - 19s 387ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 156/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 157/300\n",
      "49/49 [==============================] - 19s 386ms/step - loss: 2.3026 - accuracy: 0.1032\n",
      "Epoch 158/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 159/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49/49 [==============================] - 20s 405ms/step - loss: 2.3026 - accuracy: 0.0993\n",
      "Epoch 160/300\n",
      "49/49 [==============================] - 19s 397ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 161/300\n",
      "49/49 [==============================] - 20s 401ms/step - loss: 2.3026 - accuracy: 0.0976\n",
      "Epoch 162/300\n",
      "49/49 [==============================] - 19s 395ms/step - loss: 2.3026 - accuracy: 0.0961\n",
      "Epoch 163/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0979\n",
      "Epoch 164/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0986\n",
      "Epoch 165/300\n",
      "49/49 [==============================] - 19s 378ms/step - loss: 2.3026 - accuracy: 0.0965\n",
      "Epoch 166/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0978\n",
      "Epoch 167/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.1022\n",
      "Epoch 168/300\n",
      "49/49 [==============================] - 18s 373ms/step - loss: 2.3026 - accuracy: 0.0991\n",
      "Epoch 169/300\n",
      "49/49 [==============================] - 19s 388ms/step - loss: 2.3026 - accuracy: 0.0961\n",
      "Epoch 170/300\n",
      "49/49 [==============================] - 19s 378ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 171/300\n",
      "49/49 [==============================] - 19s 378ms/step - loss: 2.3026 - accuracy: 0.1001\n",
      "Epoch 172/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.1003\n",
      "Epoch 173/300\n",
      "49/49 [==============================] - 19s 390ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 174/300\n",
      "49/49 [==============================] - 19s 398ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 175/300\n",
      "49/49 [==============================] - 20s 413ms/step - loss: 2.3026 - accuracy: 0.1011\n",
      "Epoch 176/300\n",
      "49/49 [==============================] - 20s 406ms/step - loss: 2.3026 - accuracy: 0.1031\n",
      "Epoch 177/300\n",
      "49/49 [==============================] - 21s 427ms/step - loss: 2.3026 - accuracy: 0.0984\n",
      "Epoch 178/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0959\n",
      "Epoch 179/300\n",
      "49/49 [==============================] - 20s 401ms/step - loss: 2.3026 - accuracy: 0.0950\n",
      "Epoch 180/300\n",
      "49/49 [==============================] - 18s 379ms/step - loss: 2.3026 - accuracy: 0.0963\n",
      "Epoch 181/300\n",
      "49/49 [==============================] - 18s 368ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 182/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 183/300\n",
      "49/49 [==============================] - 20s 414ms/step - loss: 2.3026 - accuracy: 0.1000\n",
      "Epoch 184/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.0986\n",
      "Epoch 185/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 186/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.1003\n",
      "Epoch 187/300\n",
      "49/49 [==============================] - 20s 400ms/step - loss: 2.3026 - accuracy: 0.0981\n",
      "Epoch 188/300\n",
      "49/49 [==============================] - 20s 417ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 189/300\n",
      "49/49 [==============================] - 20s 412ms/step - loss: 2.3026 - accuracy: 0.0957\n",
      "Epoch 190/300\n",
      "49/49 [==============================] - 20s 418ms/step - loss: 2.3026 - accuracy: 0.1012\n",
      "Epoch 191/300\n",
      "49/49 [==============================] - 20s 401ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 192/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 193/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.0971\n",
      "Epoch 194/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 195/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0995\n",
      "Epoch 196/300\n",
      "49/49 [==============================] - 19s 395ms/step - loss: 2.3026 - accuracy: 0.1013\n",
      "Epoch 197/300\n",
      "49/49 [==============================] - 19s 375ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 198/300\n",
      "49/49 [==============================] - 19s 386ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 199/300\n",
      "49/49 [==============================] - 19s 386ms/step - loss: 2.3026 - accuracy: 0.1007\n",
      "Epoch 200/300\n",
      "49/49 [==============================] - 19s 390ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 201/300\n",
      "49/49 [==============================] - 19s 390ms/step - loss: 2.3026 - accuracy: 0.0997\n",
      "Epoch 202/300\n",
      "49/49 [==============================] - 20s 406ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 203/300\n",
      "49/49 [==============================] - 20s 400ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 204/300\n",
      "49/49 [==============================] - 19s 381ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 205/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.1015\n",
      "Epoch 206/300\n",
      "49/49 [==============================] - 18s 364ms/step - loss: 2.3026 - accuracy: 0.1009\n",
      "Epoch 207/300\n",
      "49/49 [==============================] - 18s 367ms/step - loss: 2.3026 - accuracy: 0.0989\n",
      "Epoch 208/300\n",
      "49/49 [==============================] - 18s 376ms/step - loss: 2.3026 - accuracy: 0.0982\n",
      "Epoch 209/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.0975\n",
      "Epoch 210/300\n",
      "49/49 [==============================] - 18s 376ms/step - loss: 2.3026 - accuracy: 0.1018\n",
      "Epoch 211/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.1018\n",
      "Epoch 212/300\n",
      "49/49 [==============================] - 19s 390ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 213/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.0967\n",
      "Epoch 214/300\n",
      "49/49 [==============================] - 19s 383ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 215/300\n",
      "49/49 [==============================] - 19s 393ms/step - loss: 2.3026 - accuracy: 0.0988\n",
      "Epoch 216/300\n",
      "49/49 [==============================] - 20s 411ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 217/300\n",
      "49/49 [==============================] - 19s 393ms/step - loss: 2.3026 - accuracy: 0.0989\n",
      "Epoch 218/300\n",
      "49/49 [==============================] - 19s 393ms/step - loss: 2.3026 - accuracy: 0.0975\n",
      "Epoch 219/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.1008\n",
      "Epoch 220/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0966\n",
      "Epoch 221/300\n",
      "49/49 [==============================] - 18s 368ms/step - loss: 2.3026 - accuracy: 0.1000\n",
      "Epoch 222/300\n",
      "49/49 [==============================] - 18s 370ms/step - loss: 2.3026 - accuracy: 0.0960\n",
      "Epoch 223/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.0967\n",
      "Epoch 224/300\n",
      "49/49 [==============================] - 19s 379ms/step - loss: 2.3026 - accuracy: 0.0986\n",
      "Epoch 225/300\n",
      "49/49 [==============================] - 18s 364ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 226/300\n",
      "49/49 [==============================] - 19s 384ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 227/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.1014\n",
      "Epoch 228/300\n",
      "49/49 [==============================] - 19s 384ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 229/300\n",
      "49/49 [==============================] - 19s 389ms/step - loss: 2.3026 - accuracy: 0.1007\n",
      "Epoch 230/300\n",
      "49/49 [==============================] - 19s 393ms/step - loss: 2.3026 - accuracy: 0.0997\n",
      "Epoch 231/300\n",
      "49/49 [==============================] - 20s 405ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 232/300\n",
      "49/49 [==============================] - 20s 405ms/step - loss: 2.3026 - accuracy: 0.0979\n",
      "Epoch 233/300\n",
      "49/49 [==============================] - 19s 391ms/step - loss: 2.3026 - accuracy: 0.0963\n",
      "Epoch 234/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0989\n",
      "Epoch 235/300\n",
      "49/49 [==============================] - 19s 378ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 236/300\n",
      "49/49 [==============================] - 18s 372ms/step - loss: 2.3026 - accuracy: 0.1001\n",
      "Epoch 237/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 238/300\n",
      "49/49 [==============================] - 18s 376ms/step - loss: 2.3026 - accuracy: 0.0995\n",
      "Epoch 239/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.0973\n",
      "Epoch 240/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.0985\n",
      "Epoch 241/300\n",
      "49/49 [==============================] - 19s 381ms/step - loss: 2.3026 - accuracy: 0.1014\n",
      "Epoch 242/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0981\n",
      "Epoch 243/300\n",
      "49/49 [==============================] - 19s 387ms/step - loss: 2.3026 - accuracy: 0.0968\n",
      "Epoch 244/300\n",
      "49/49 [==============================] - 19s 386ms/step - loss: 2.3026 - accuracy: 0.1007\n",
      "Epoch 245/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 246/300\n",
      "49/49 [==============================] - 19s 386ms/step - loss: 2.3026 - accuracy: 0.0981\n",
      "Epoch 247/300\n",
      "49/49 [==============================] - 20s 402ms/step - loss: 2.3026 - accuracy: 0.0980\n",
      "Epoch 248/300\n",
      "49/49 [==============================] - 20s 399ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 249/300\n",
      "49/49 [==============================] - 21s 422ms/step - loss: 2.3026 - accuracy: 0.0984\n",
      "Epoch 250/300\n",
      "49/49 [==============================] - 20s 415ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 251/300\n",
      "49/49 [==============================] - 19s 398ms/step - loss: 2.3026 - accuracy: 0.0980\n",
      "Epoch 252/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.0976\n",
      "Epoch 253/300\n",
      "49/49 [==============================] - 18s 375ms/step - loss: 2.3026 - accuracy: 0.0976\n",
      "Epoch 254/300\n",
      "49/49 [==============================] - 18s 378ms/step - loss: 2.3026 - accuracy: 0.0971\n",
      "Epoch 255/300\n",
      "49/49 [==============================] - 18s 368ms/step - loss: 2.3026 - accuracy: 0.0956\n",
      "Epoch 256/300\n",
      "49/49 [==============================] - 18s 366ms/step - loss: 2.3026 - accuracy: 0.0974\n",
      "Epoch 257/300\n",
      "49/49 [==============================] - 18s 362ms/step - loss: 2.3026 - accuracy: 0.0980\n",
      "Epoch 258/300\n",
      "49/49 [==============================] - 19s 385ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 259/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 260/300\n",
      "49/49 [==============================] - 19s 397ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 261/300\n",
      "49/49 [==============================] - 19s 397ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 262/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0958\n",
      "Epoch 263/300\n",
      "49/49 [==============================] - 20s 404ms/step - loss: 2.3026 - accuracy: 0.1003\n",
      "Epoch 264/300\n",
      "49/49 [==============================] - 19s 392ms/step - loss: 2.3026 - accuracy: 0.0991\n",
      "Epoch 265/300\n",
      "49/49 [==============================] - 19s 397ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 266/300\n",
      "49/49 [==============================] - 19s 389ms/step - loss: 2.3026 - accuracy: 0.0964\n",
      "Epoch 267/300\n",
      "49/49 [==============================] - 18s 368ms/step - loss: 2.3026 - accuracy: 0.0999\n",
      "Epoch 268/300\n",
      "49/49 [==============================] - 17s 356ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 269/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.0961\n",
      "Epoch 270/300\n",
      "49/49 [==============================] - 18s 367ms/step - loss: 2.3026 - accuracy: 0.0977\n",
      "Epoch 271/300\n",
      "49/49 [==============================] - 18s 365ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 272/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.1034\n",
      "Epoch 273/300\n",
      "49/49 [==============================] - 18s 367ms/step - loss: 2.3026 - accuracy: 0.1001\n",
      "Epoch 274/300\n",
      "49/49 [==============================] - 19s 384ms/step - loss: 2.3026 - accuracy: 0.0998\n",
      "Epoch 275/300\n",
      "49/49 [==============================] - 19s 385ms/step - loss: 2.3026 - accuracy: 0.0965\n",
      "Epoch 276/300\n",
      "49/49 [==============================] - 19s 380ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 277/300\n",
      "49/49 [==============================] - 19s 394ms/step - loss: 2.3026 - accuracy: 0.1007\n",
      "Epoch 278/300\n",
      "49/49 [==============================] - 19s 398ms/step - loss: 2.3026 - accuracy: 0.0982\n",
      "Epoch 279/300\n",
      "49/49 [==============================] - 20s 417ms/step - loss: 2.3026 - accuracy: 0.0991\n",
      "Epoch 280/300\n",
      "49/49 [==============================] - 20s 401ms/step - loss: 2.3026 - accuracy: 0.0992\n",
      "Epoch 281/300\n",
      "49/49 [==============================] - 19s 393ms/step - loss: 2.3026 - accuracy: 0.1012\n",
      "Epoch 282/300\n",
      "49/49 [==============================] - 19s 383ms/step - loss: 2.3026 - accuracy: 0.0987\n",
      "Epoch 283/300\n",
      "49/49 [==============================] - 18s 373ms/step - loss: 2.3026 - accuracy: 0.0990\n",
      "Epoch 284/300\n",
      "49/49 [==============================] - 18s 377ms/step - loss: 2.3026 - accuracy: 0.1002\n",
      "Epoch 285/300\n",
      "49/49 [==============================] - 18s 363ms/step - loss: 2.3026 - accuracy: 0.1021\n",
      "Epoch 286/300\n",
      "49/49 [==============================] - 18s 374ms/step - loss: 2.3026 - accuracy: 0.0989\n",
      "Epoch 287/300\n",
      "49/49 [==============================] - 18s 369ms/step - loss: 2.3026 - accuracy: 0.1014\n",
      "Epoch 288/300\n",
      "49/49 [==============================] - 19s 388ms/step - loss: 2.3026 - accuracy: 0.1005\n",
      "Epoch 289/300\n",
      "49/49 [==============================] - 19s 381ms/step - loss: 2.3026 - accuracy: 0.0989\n",
      "Epoch 290/300\n",
      "49/49 [==============================] - 19s 391ms/step - loss: 2.3026 - accuracy: 0.0983\n",
      "Epoch 291/300\n",
      "49/49 [==============================] - 19s 396ms/step - loss: 2.3026 - accuracy: 0.0956\n",
      "Epoch 292/300\n",
      "49/49 [==============================] - 19s 395ms/step - loss: 2.3026 - accuracy: 0.1009\n",
      "Epoch 293/300\n",
      "49/49 [==============================] - 19s 398ms/step - loss: 2.3026 - accuracy: 0.0977\n",
      "Epoch 294/300\n",
      "49/49 [==============================] - 19s 394ms/step - loss: 2.3026 - accuracy: 0.1010\n",
      "Epoch 295/300\n",
      "49/49 [==============================] - 19s 392ms/step - loss: 2.3026 - accuracy: 0.0990\n",
      "Epoch 296/300\n",
      "49/49 [==============================] - 19s 385ms/step - loss: 2.3026 - accuracy: 0.0991\n",
      "Epoch 297/300\n",
      "49/49 [==============================] - 18s 371ms/step - loss: 2.3026 - accuracy: 0.0994\n",
      "Epoch 298/300\n",
      "49/49 [==============================] - 18s 368ms/step - loss: 2.3026 - accuracy: 0.0977\n",
      "Epoch 299/300\n",
      "49/49 [==============================] - 18s 369ms/step - loss: 2.3026 - accuracy: 0.0996\n",
      "Epoch 300/300\n",
      "49/49 [==============================] - 17s 339ms/step - loss: 2.3026 - accuracy: 0.0971\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "max_pooling2d_1 (MaxPooling2 (None, 15, 15, 48)        0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 10800)             0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1024)              11060224  \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                10250     \n",
      "=================================================================\n",
      "Total params: 11,070,474\n",
      "Trainable params: 11,070,474\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Initialize the He initializer\n",
    "initializer = tf.keras.initializers.HeNormal()\n",
    "\n",
    "model_1 = Sequential()\n",
    "\n",
    "model_1.add(Input(shape=(30, 30, 48)))\n",
    "model_1.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model_1.add(Flatten())\n",
    "model_1.add(Dense(1024, activation='relu', kernel_initializer=initializer))\n",
    "model_1.add(Dense(10, activation='softmax', kernel_initializer=initializer))\n",
    "\n",
    "model_1.compile(optimizer='adam', \n",
    "               loss='sparse_categorical_crossentropy', \n",
    "               metrics = ['accuracy'])\n",
    "\n",
    "history_data1 = model_1.fit(x=train_data, \n",
    "                                     y=train_labels, \n",
    "                                     epochs=300,\n",
    "                                     batch_size=1024)\n",
    "\n",
    "model_1.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Experiment 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 13.5727 - accuracy: 0.1017\n",
      "Epoch 2/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 2.3177 - accuracy: 0.1069\n",
      "Epoch 3/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 2.2969 - accuracy: 0.1079\n",
      "Epoch 4/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 2.2834 - accuracy: 0.1179\n",
      "Epoch 5/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 2.2742 - accuracy: 0.1182\n",
      "Epoch 6/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 2.2562 - accuracy: 0.1277\n",
      "Epoch 7/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 2.2401 - accuracy: 0.1376\n",
      "Epoch 8/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 2.2226 - accuracy: 0.1426\n",
      "Epoch 9/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 2.1993 - accuracy: 0.1510\n",
      "Epoch 10/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 2.1662 - accuracy: 0.1676\n",
      "Epoch 11/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 2.1341 - accuracy: 0.1808\n",
      "Epoch 12/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 2.0992 - accuracy: 0.1951\n",
      "Epoch 13/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 2.0606 - accuracy: 0.2138\n",
      "Epoch 14/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 2.0156 - accuracy: 0.2301\n",
      "Epoch 15/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.9765 - accuracy: 0.2453\n",
      "Epoch 16/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.9344 - accuracy: 0.2624\n",
      "Epoch 17/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 1.8971 - accuracy: 0.2753\n",
      "Epoch 18/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 1.8457 - accuracy: 0.2945\n",
      "Epoch 19/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.8024 - accuracy: 0.3132\n",
      "Epoch 20/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.7550 - accuracy: 0.3332\n",
      "Epoch 21/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.7156 - accuracy: 0.3493\n",
      "Epoch 22/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 1.6769 - accuracy: 0.3636\n",
      "Epoch 23/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 1.6189 - accuracy: 0.3826\n",
      "Epoch 24/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.5868 - accuracy: 0.3945\n",
      "Epoch 25/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.5312 - accuracy: 0.4164\n",
      "Epoch 26/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 1.4978 - accuracy: 0.4314\n",
      "Epoch 27/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 1.4672 - accuracy: 0.4431\n",
      "Epoch 28/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 1.4318 - accuracy: 0.4545\n",
      "Epoch 29/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 1.3802 - accuracy: 0.4713\n",
      "Epoch 30/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.3497 - accuracy: 0.4836\n",
      "Epoch 31/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.3008 - accuracy: 0.5043\n",
      "Epoch 32/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 1.2789 - accuracy: 0.5117\n",
      "Epoch 33/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 1.2328 - accuracy: 0.5264\n",
      "Epoch 34/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 1.1915 - accuracy: 0.5439\n",
      "Epoch 35/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.1639 - accuracy: 0.5557\n",
      "Epoch 36/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 1.1313 - accuracy: 0.5682\n",
      "Epoch 37/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 1.0959 - accuracy: 0.5787\n",
      "Epoch 38/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 1.0631 - accuracy: 0.5953\n",
      "Epoch 39/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 1.0367 - accuracy: 0.6052\n",
      "Epoch 40/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 1.0063 - accuracy: 0.6175\n",
      "Epoch 41/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.9927 - accuracy: 0.6227\n",
      "Epoch 42/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.9579 - accuracy: 0.6359\n",
      "Epoch 43/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.9283 - accuracy: 0.6483\n",
      "Epoch 44/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.9035 - accuracy: 0.6617\n",
      "Epoch 45/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.8653 - accuracy: 0.6719\n",
      "Epoch 46/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.8399 - accuracy: 0.6837\n",
      "Epoch 47/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.8268 - accuracy: 0.6896\n",
      "Epoch 48/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.8009 - accuracy: 0.6987\n",
      "Epoch 49/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.7837 - accuracy: 0.7042\n",
      "Epoch 50/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.7661 - accuracy: 0.7122\n",
      "Epoch 51/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.7412 - accuracy: 0.7208\n",
      "Epoch 52/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.7281 - accuracy: 0.7299\n",
      "Epoch 53/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.6933 - accuracy: 0.7395\n",
      "Epoch 54/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.6794 - accuracy: 0.7471\n",
      "Epoch 55/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.6815 - accuracy: 0.7460\n",
      "Epoch 56/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.6525 - accuracy: 0.7594\n",
      "Epoch 57/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.6401 - accuracy: 0.7639\n",
      "Epoch 58/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.6112 - accuracy: 0.7724\n",
      "Epoch 59/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.5977 - accuracy: 0.7761\n",
      "Epoch 60/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.5975 - accuracy: 0.7773\n",
      "Epoch 61/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.5811 - accuracy: 0.7846\n",
      "Epoch 62/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.5611 - accuracy: 0.7932\n",
      "Epoch 63/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.5421 - accuracy: 0.8007\n",
      "Epoch 64/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.5473 - accuracy: 0.7969\n",
      "Epoch 65/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.5369 - accuracy: 0.8015\n",
      "Epoch 66/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.5201 - accuracy: 0.8077\n",
      "Epoch 67/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.5032 - accuracy: 0.8133\n",
      "Epoch 68/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.4938 - accuracy: 0.8169\n",
      "Epoch 69/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.4937 - accuracy: 0.8176\n",
      "Epoch 70/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.4820 - accuracy: 0.8233\n",
      "Epoch 71/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.4561 - accuracy: 0.8313\n",
      "Epoch 72/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.4760 - accuracy: 0.8255\n",
      "Epoch 73/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.4593 - accuracy: 0.8326\n",
      "Epoch 74/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.4448 - accuracy: 0.8378\n",
      "Epoch 75/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.4211 - accuracy: 0.8470\n",
      "Epoch 76/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.4220 - accuracy: 0.8470\n",
      "Epoch 77/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.4179 - accuracy: 0.8488\n",
      "Epoch 78/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.3998 - accuracy: 0.8547\n",
      "Epoch 79/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.4030 - accuracy: 0.8541\n",
      "Epoch 80/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3956 - accuracy: 0.8566\n",
      "Epoch 81/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3842 - accuracy: 0.8602\n",
      "Epoch 82/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.3747 - accuracy: 0.8642\n",
      "Epoch 83/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49/49 [==============================] - 63s 1s/step - loss: 0.3738 - accuracy: 0.8637\n",
      "Epoch 84/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3690 - accuracy: 0.8665\n",
      "Epoch 85/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3520 - accuracy: 0.8731\n",
      "Epoch 86/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.3610 - accuracy: 0.8719\n",
      "Epoch 87/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.3616 - accuracy: 0.8727\n",
      "Epoch 88/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3507 - accuracy: 0.8731\n",
      "Epoch 89/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.3448 - accuracy: 0.8775\n",
      "Epoch 90/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3335 - accuracy: 0.8813\n",
      "Epoch 91/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.3370 - accuracy: 0.8799\n",
      "Epoch 92/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.3278 - accuracy: 0.8810\n",
      "Epoch 93/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3139 - accuracy: 0.8868\n",
      "Epoch 94/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3017 - accuracy: 0.8916\n",
      "Epoch 95/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3058 - accuracy: 0.8901\n",
      "Epoch 96/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.3053 - accuracy: 0.8914\n",
      "Epoch 97/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.3009 - accuracy: 0.8928\n",
      "Epoch 98/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.3012 - accuracy: 0.8941\n",
      "Epoch 99/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.3059 - accuracy: 0.8908\n",
      "Epoch 100/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2911 - accuracy: 0.8969\n",
      "Epoch 101/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2895 - accuracy: 0.8968\n",
      "Epoch 102/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2835 - accuracy: 0.8990\n",
      "Epoch 103/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2750 - accuracy: 0.9017\n",
      "Epoch 104/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2744 - accuracy: 0.9033\n",
      "Epoch 105/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2812 - accuracy: 0.9018\n",
      "Epoch 106/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2689 - accuracy: 0.9051\n",
      "Epoch 107/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2716 - accuracy: 0.9031\n",
      "Epoch 108/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2686 - accuracy: 0.9028\n",
      "Epoch 109/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2573 - accuracy: 0.9091\n",
      "Epoch 110/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2598 - accuracy: 0.9112\n",
      "Epoch 111/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.2590 - accuracy: 0.9105\n",
      "Epoch 112/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2471 - accuracy: 0.9128\n",
      "Epoch 113/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2425 - accuracy: 0.9151\n",
      "Epoch 114/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2402 - accuracy: 0.9157\n",
      "Epoch 115/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2423 - accuracy: 0.9148\n",
      "Epoch 116/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.2389 - accuracy: 0.9161\n",
      "Epoch 117/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2245 - accuracy: 0.9205\n",
      "Epoch 118/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2268 - accuracy: 0.9204\n",
      "Epoch 119/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2356 - accuracy: 0.9207\n",
      "Epoch 120/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.2318 - accuracy: 0.9190\n",
      "Epoch 121/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2280 - accuracy: 0.9184\n",
      "Epoch 122/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2210 - accuracy: 0.9221\n",
      "Epoch 123/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2212 - accuracy: 0.9221\n",
      "Epoch 124/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2193 - accuracy: 0.9224\n",
      "Epoch 125/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2126 - accuracy: 0.9266\n",
      "Epoch 126/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.2129 - accuracy: 0.9243\n",
      "Epoch 127/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2160 - accuracy: 0.9250\n",
      "Epoch 128/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2047 - accuracy: 0.9265\n",
      "Epoch 129/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2137 - accuracy: 0.9270\n",
      "Epoch 130/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1960 - accuracy: 0.9295\n",
      "Epoch 131/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.2025 - accuracy: 0.9306\n",
      "Epoch 132/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.2075 - accuracy: 0.9275\n",
      "Epoch 133/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2008 - accuracy: 0.9313\n",
      "Epoch 134/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.2042 - accuracy: 0.9286\n",
      "Epoch 135/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1932 - accuracy: 0.9328\n",
      "Epoch 136/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1958 - accuracy: 0.9324\n",
      "Epoch 137/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1999 - accuracy: 0.9321\n",
      "Epoch 138/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1928 - accuracy: 0.9330\n",
      "Epoch 139/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1979 - accuracy: 0.9337\n",
      "Epoch 140/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1933 - accuracy: 0.9315\n",
      "Epoch 141/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1928 - accuracy: 0.9338\n",
      "Epoch 142/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1747 - accuracy: 0.9411\n",
      "Epoch 143/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1806 - accuracy: 0.9365\n",
      "Epoch 144/300\n",
      "49/49 [==============================] - 65s 1s/step - loss: 0.1812 - accuracy: 0.9359\n",
      "Epoch 145/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1833 - accuracy: 0.9383\n",
      "Epoch 146/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1857 - accuracy: 0.9362\n",
      "Epoch 147/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1873 - accuracy: 0.9352\n",
      "Epoch 148/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1754 - accuracy: 0.9381\n",
      "Epoch 149/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1749 - accuracy: 0.9405\n",
      "Epoch 150/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1827 - accuracy: 0.9385\n",
      "Epoch 151/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1641 - accuracy: 0.9443\n",
      "Epoch 152/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1604 - accuracy: 0.9443\n",
      "Epoch 153/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1709 - accuracy: 0.9400\n",
      "Epoch 154/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1631 - accuracy: 0.9443\n",
      "Epoch 155/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1684 - accuracy: 0.9401\n",
      "Epoch 156/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1678 - accuracy: 0.9413\n",
      "Epoch 157/300\n",
      "49/49 [==============================] - 65s 1s/step - loss: 0.1620 - accuracy: 0.9445\n",
      "Epoch 158/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1650 - accuracy: 0.9444\n",
      "Epoch 159/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1695 - accuracy: 0.9434\n",
      "Epoch 160/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1665 - accuracy: 0.9428\n",
      "Epoch 161/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1584 - accuracy: 0.9448\n",
      "Epoch 162/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1706 - accuracy: 0.9408\n",
      "Epoch 163/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1631 - accuracy: 0.9439\n",
      "Epoch 164/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49/49 [==============================] - 59s 1s/step - loss: 0.1522 - accuracy: 0.9478\n",
      "Epoch 165/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1529 - accuracy: 0.9458\n",
      "Epoch 166/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1596 - accuracy: 0.9462\n",
      "Epoch 167/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1639 - accuracy: 0.9437\n",
      "Epoch 168/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1681 - accuracy: 0.9424\n",
      "Epoch 169/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1535 - accuracy: 0.9470\n",
      "Epoch 170/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1565 - accuracy: 0.9459\n",
      "Epoch 171/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1453 - accuracy: 0.9510\n",
      "Epoch 172/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1453 - accuracy: 0.9507\n",
      "Epoch 173/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1406 - accuracy: 0.9499\n",
      "Epoch 174/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1452 - accuracy: 0.9494\n",
      "Epoch 175/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1393 - accuracy: 0.9521\n",
      "Epoch 176/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1410 - accuracy: 0.9504\n",
      "Epoch 177/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1446 - accuracy: 0.9512\n",
      "Epoch 178/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1474 - accuracy: 0.9505\n",
      "Epoch 179/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1482 - accuracy: 0.9515\n",
      "Epoch 180/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1430 - accuracy: 0.9519\n",
      "Epoch 181/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1412 - accuracy: 0.9522\n",
      "Epoch 182/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1424 - accuracy: 0.9507\n",
      "Epoch 183/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1501 - accuracy: 0.9494\n",
      "Epoch 184/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1409 - accuracy: 0.9527\n",
      "Epoch 185/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1388 - accuracy: 0.9538\n",
      "Epoch 186/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1350 - accuracy: 0.9544\n",
      "Epoch 187/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1406 - accuracy: 0.9516\n",
      "Epoch 188/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1265 - accuracy: 0.9569\n",
      "Epoch 189/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1329 - accuracy: 0.9546\n",
      "Epoch 190/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1309 - accuracy: 0.9552\n",
      "Epoch 191/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1269 - accuracy: 0.9566\n",
      "Epoch 192/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1258 - accuracy: 0.9572\n",
      "Epoch 193/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1245 - accuracy: 0.9590\n",
      "Epoch 194/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1305 - accuracy: 0.9541\n",
      "Epoch 195/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1241 - accuracy: 0.9572\n",
      "Epoch 196/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1206 - accuracy: 0.9595\n",
      "Epoch 197/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1260 - accuracy: 0.9576\n",
      "Epoch 198/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1235 - accuracy: 0.9576\n",
      "Epoch 199/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1278 - accuracy: 0.9573\n",
      "Epoch 200/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1300 - accuracy: 0.9559\n",
      "Epoch 201/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1246 - accuracy: 0.9572\n",
      "Epoch 202/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1389 - accuracy: 0.9554\n",
      "Epoch 203/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1350 - accuracy: 0.9546\n",
      "Epoch 204/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1338 - accuracy: 0.9554\n",
      "Epoch 205/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1304 - accuracy: 0.9566\n",
      "Epoch 206/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1316 - accuracy: 0.9552\n",
      "Epoch 207/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1257 - accuracy: 0.9581\n",
      "Epoch 208/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1216 - accuracy: 0.9569\n",
      "Epoch 209/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1247 - accuracy: 0.9580\n",
      "Epoch 210/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1180 - accuracy: 0.9598\n",
      "Epoch 211/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1226 - accuracy: 0.9575\n",
      "Epoch 212/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1216 - accuracy: 0.9592\n",
      "Epoch 213/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1205 - accuracy: 0.9591\n",
      "Epoch 214/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1226 - accuracy: 0.9607\n",
      "Epoch 215/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1169 - accuracy: 0.9603\n",
      "Epoch 216/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1261 - accuracy: 0.9592\n",
      "Epoch 217/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1203 - accuracy: 0.9583\n",
      "Epoch 218/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1210 - accuracy: 0.9595\n",
      "Epoch 219/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1210 - accuracy: 0.9588\n",
      "Epoch 220/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1296 - accuracy: 0.9580\n",
      "Epoch 221/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1175 - accuracy: 0.9617\n",
      "Epoch 222/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1086 - accuracy: 0.9616\n",
      "Epoch 223/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1079 - accuracy: 0.9636\n",
      "Epoch 224/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1138 - accuracy: 0.9623\n",
      "Epoch 225/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1076 - accuracy: 0.9618\n",
      "Epoch 226/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1123 - accuracy: 0.9628\n",
      "Epoch 227/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1111 - accuracy: 0.9618\n",
      "Epoch 228/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1150 - accuracy: 0.9624\n",
      "Epoch 229/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1096 - accuracy: 0.9629\n",
      "Epoch 230/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1076 - accuracy: 0.9629\n",
      "Epoch 231/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0996 - accuracy: 0.9651\n",
      "Epoch 232/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1030 - accuracy: 0.9643\n",
      "Epoch 233/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1061 - accuracy: 0.9643\n",
      "Epoch 234/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1028 - accuracy: 0.9653\n",
      "Epoch 235/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1035 - accuracy: 0.9635\n",
      "Epoch 236/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1073 - accuracy: 0.9638\n",
      "Epoch 237/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1029 - accuracy: 0.9654\n",
      "Epoch 238/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1053 - accuracy: 0.9641\n",
      "Epoch 239/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1004 - accuracy: 0.9651\n",
      "Epoch 240/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1046 - accuracy: 0.9635\n",
      "Epoch 241/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0931 - accuracy: 0.9676\n",
      "Epoch 242/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1007 - accuracy: 0.9658\n",
      "Epoch 243/300\n",
      "49/49 [==============================] - 64s 1s/step - loss: 0.1001 - accuracy: 0.9658\n",
      "Epoch 244/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1019 - accuracy: 0.9660\n",
      "Epoch 245/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49/49 [==============================] - 60s 1s/step - loss: 0.1061 - accuracy: 0.9663\n",
      "Epoch 246/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1120 - accuracy: 0.9636\n",
      "Epoch 247/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1097 - accuracy: 0.9632\n",
      "Epoch 248/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1072 - accuracy: 0.9642\n",
      "Epoch 249/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.1066 - accuracy: 0.9635\n",
      "Epoch 250/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.1081 - accuracy: 0.9645\n",
      "Epoch 251/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1037 - accuracy: 0.9648\n",
      "Epoch 252/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1086 - accuracy: 0.9637\n",
      "Epoch 253/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.1022 - accuracy: 0.9650\n",
      "Epoch 254/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.0975 - accuracy: 0.9657\n",
      "Epoch 255/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.0977 - accuracy: 0.9667\n",
      "Epoch 256/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0908 - accuracy: 0.9687\n",
      "Epoch 257/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.1003 - accuracy: 0.9665\n",
      "Epoch 258/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.1012 - accuracy: 0.9653\n",
      "Epoch 259/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.0992 - accuracy: 0.9664\n",
      "Epoch 260/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0970 - accuracy: 0.9662\n",
      "Epoch 261/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0918 - accuracy: 0.9684\n",
      "Epoch 262/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.0937 - accuracy: 0.9694\n",
      "Epoch 263/300\n",
      "49/49 [==============================] - 65s 1s/step - loss: 0.0920 - accuracy: 0.9680\n",
      "Epoch 264/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.0926 - accuracy: 0.9689\n",
      "Epoch 265/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0890 - accuracy: 0.9701\n",
      "Epoch 266/300\n",
      "49/49 [==============================] - 60s 1s/step - loss: 0.0912 - accuracy: 0.9690\n",
      "Epoch 267/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.0896 - accuracy: 0.9702\n",
      "Epoch 268/300\n",
      "49/49 [==============================] - 63s 1s/step - loss: 0.0903 - accuracy: 0.9681\n",
      "Epoch 269/300\n",
      "49/49 [==============================] - 62s 1s/step - loss: 0.0894 - accuracy: 0.9695\n",
      "Epoch 270/300\n",
      "49/49 [==============================] - 59s 1s/step - loss: 0.0887 - accuracy: 0.9700\n",
      "Epoch 271/300\n",
      "49/49 [==============================] - 61s 1s/step - loss: 0.0906 - accuracy: 0.9689\n",
      "Epoch 272/300\n",
      "49/49 [==============================] - 82s 2s/step - loss: 0.0903 - accuracy: 0.9690\n",
      "Epoch 273/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0875 - accuracy: 0.9698\n",
      "Epoch 274/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0902 - accuracy: 0.9703\n",
      "Epoch 275/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0899 - accuracy: 0.9696\n",
      "Epoch 276/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0897 - accuracy: 0.9701\n",
      "Epoch 277/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0922 - accuracy: 0.9692\n",
      "Epoch 278/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0942 - accuracy: 0.9674\n",
      "Epoch 279/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0878 - accuracy: 0.9698\n",
      "Epoch 280/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0914 - accuracy: 0.9694\n",
      "Epoch 281/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0929 - accuracy: 0.9704\n",
      "Epoch 282/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0937 - accuracy: 0.9700\n",
      "Epoch 283/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0895 - accuracy: 0.9689\n",
      "Epoch 284/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0833 - accuracy: 0.9717\n",
      "Epoch 285/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0834 - accuracy: 0.9723\n",
      "Epoch 286/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0910 - accuracy: 0.9689\n",
      "Epoch 287/300\n",
      "49/49 [==============================] - 96s 2s/step - loss: 0.0910 - accuracy: 0.9689\n",
      "Epoch 288/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0909 - accuracy: 0.9707\n",
      "Epoch 289/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0935 - accuracy: 0.9693\n",
      "Epoch 290/300\n",
      "49/49 [==============================] - 99s 2s/step - loss: 0.0893 - accuracy: 0.9709\n",
      "Epoch 291/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0890 - accuracy: 0.9701\n",
      "Epoch 292/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0883 - accuracy: 0.9700\n",
      "Epoch 293/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0885 - accuracy: 0.9703\n",
      "Epoch 294/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0863 - accuracy: 0.9697\n",
      "Epoch 295/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0866 - accuracy: 0.9714\n",
      "Epoch 296/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0827 - accuracy: 0.9718\n",
      "Epoch 297/300\n",
      "49/49 [==============================] - 98s 2s/step - loss: 0.0890 - accuracy: 0.9705\n",
      "Epoch 298/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0840 - accuracy: 0.9715\n",
      "Epoch 299/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0873 - accuracy: 0.9698\n",
      "Epoch 300/300\n",
      "49/49 [==============================] - 97s 2s/step - loss: 0.0820 - accuracy: 0.9728\n",
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "batch_normalization (BatchNo (None, 30, 30, 48)        192       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 15, 15, 48)        0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 10800)             0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 1024)              11060224  \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 10)                10250     \n",
      "=================================================================\n",
      "Total params: 11,070,666\n",
      "Trainable params: 11,070,570\n",
      "Non-trainable params: 96\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.layers import BatchNormalization, Dropout\n",
    "\n",
    "model_2 = Sequential()\n",
    "\n",
    "model_2.add(Input(shape=(30, 30, 48)))\n",
    "model_2.add(BatchNormalization())\n",
    "model_2.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model_2.add(Flatten())\n",
    "model_2.add(Dense(1024, activation='relu', kernel_initializer=initializer))\n",
    "model_2.add(Dropout(0.2))\n",
    "model_2.add(Dense(10, activation='softmax', kernel_initializer=initializer))\n",
    "\n",
    "model_2.compile(optimizer='adam', \n",
    "               loss='sparse_categorical_crossentropy', \n",
    "               metrics = ['accuracy'])\n",
    "\n",
    "history_data2 = model_2.fit(x=train_data, \n",
    "                                     y=train_labels, \n",
    "                                     epochs=300, \n",
    "                                     batch_size=1024)\n",
    "\n",
    "model_2.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. EVALUATION"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x24a23479d00>"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(history_data1.history['loss'], label = \"train_loss\")\n",
    "plt.xlabel('iteration')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x24a4241a130>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(history_data1.history['accuracy'], label = \"train_accuracy\")\n",
    "plt.xlabel('iteration')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x24a424edbb0>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(history_data2.history['loss'], label = \"train_loss\")\n",
    "plt.xlabel('iteration')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x24a425674c0>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(history_data2.history['accuracy'], label = \"train_accuracy\")\n",
    "plt.xlabel('iteration')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. SAVE MODELS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_1.save('C://Users/MINH LE/Desktop/TraditionalvsCNN/models/Experiment 1/Exp1-CIFAR-TraditionalFilter.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_2.save('C://Users/MINH LE/Desktop/TraditionalvsCNN/models/Experiment 2/Exp2-CIFAR-TraditionalFilter.h5')"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}