AM_DLFC.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "W2UwnkTSboeP",
    "outputId": "9880acb0-957c-40eb-9506-2d8d2f943e0f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.9.0\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "print(tf.__version__)\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras import optimizers\n",
    "from keras import backend as K\n",
    "import keras\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.pyplot import figure\n",
    "from tensorflow.keras.utils import plot_model\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "import numpy as np\n",
    "import pickle\n",
    "from sklearn.utils import shuffle\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Import necessary items from Keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Activation, Dropout, UpSampling2D\n",
    "from keras.layers import Conv2DTranspose, Conv2D, MaxPooling2D\n",
    "#from keras.layers.normalization import BatchNormalization\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras import regularizers\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras.layers import Activation,Input,BatchNormalization,Add, AveragePooling2D ,Flatten,ReLU ,Add,Dense,Reshape,Multiply, Dropout, UpSampling2D,Conv2D,MaxPooling2D,concatenate,add,GlobalAveragePooling2D,GlobalMaxPool2D,Multiply"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "xsVs0965b8X8"
   },
   "outputs": [],
   "source": [
    "import hdf5storage\n",
    "mat = hdf5storage.loadmat('D:/Projects/Anni apu/KU_har_time_freq_spectrogram_SP.mat')\n",
    "data= mat['SP']\n",
    "del mat\n",
    "img_rows, img_cols = 8, 129\n",
    "Channels = 12"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "VFbr0BhIfBZQ",
    "outputId": "aaf6f2ed-6ff1-4582-ddbe-3dde1f835264"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "18\n"
     ]
    }
   ],
   "source": [
    "df=pd.read_csv(\"D:/Projects/Anni apu/kuhar_labels.csv\",header=None)\n",
    "dff=df.values\n",
    "y = dff[:,0]\n",
    "num_classes = len(set(y))\n",
    "print(num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "LrVQkdCfb8Vo",
    "outputId": "e7e2346f-218e-46cd-f7da-afdc410a2af6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20750, 8, 129, 6)\n"
     ]
    }
   ],
   "source": [
    "x= np.array(data[: ,:, :, 0:6], dtype=np.float32)\n",
    "#del data\n",
    "print(x.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "nPF-hRZfb8Mx",
    "outputId": "f6ede606-e289-4fcb-e73f-3ecb8001f20f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15562, 8, 129, 6)\n",
      "(5188, 8, 129, 6)\n",
      "17\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test= train_test_split(x,y,test_size=0.25,random_state=9,stratify=y)\n",
    "print(X_train.shape)\n",
    "print(X_test.shape)\n",
    "print(max(y_train))\n",
    "del x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15562, 18)\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.utils import to_categorical\n",
    "y_train= to_categorical(y_train)\n",
    "y_test= to_categorical(y_test)\n",
    "print(y_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5188, 18)\n"
     ]
    }
   ],
   "source": [
    "print(y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow.keras.backend as K\n",
    "\n",
    "def dsc(y_true, y_pred, smooth=1):\n",
    "    y_true_f = K.flatten(y_true)\n",
    "    y_pred_f = K.flatten(y_pred)\n",
    "    intersection = tf.reduce_sum(y_true_f * y_pred_f)\n",
    "    return (2. * intersection + smooth) / (tf.reduce_sum(y_true_f) + tf.reduce_sum(y_pred_f) + smooth)\n",
    "\n",
    "\n",
    "def dice_loss(y_true, y_pred):\n",
    "    return 1.0 - dsc(y_true, y_pred)\n",
    "\n",
    "\n",
    "\n",
    "def IOU(y_true, y_pred):\n",
    "\n",
    "    y_true = K.flatten(y_true)\n",
    "    y_pred = K.flatten(y_pred)\n",
    "\n",
    "    thresh = 0.5\n",
    "\n",
    "    y_true = K.cast(K.greater_equal(y_true, thresh), 'float32')\n",
    "    y_pred = K.cast(K.greater_equal(y_pred, thresh), 'float32')\n",
    "\n",
    "    union = K.sum(K.maximum(y_true, y_pred)) + K.epsilon()\n",
    "    intersection = K.sum(K.minimum(y_true, y_pred)) + K.epsilon()\n",
    "\n",
    "    iou = intersection/union\n",
    "\n",
    "    return iou\n",
    "\n",
    "def recall_m(y_true, y_pred):\n",
    "    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
    "    possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
    "    recall = true_positives / (possible_positives + K.epsilon())\n",
    "    return recall\n",
    "\n",
    "def precision_m(y_true, y_pred):\n",
    "    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
    "    predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
    "    precision = true_positives / (predicted_positives + K.epsilon())\n",
    "    return precision\n",
    "\n",
    "def f1_m(y_true, y_pred):\n",
    "    precision = precision_m(y_true, y_pred)\n",
    "    recall = recall_m(y_true, y_pred)\n",
    "    return 2*((precision*recall)/(precision+recall+K.epsilon()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def CBAM (x,r):\n",
    "    \n",
    "    ###### Channel Attention ##########\n",
    "    f=x.shape[-1]\n",
    "    a_pool = GlobalAveragePooling2D()(x)\n",
    "    m_pool = GlobalMaxPool2D()(x) \n",
    "    Dense_layer1 = Dense (f // r , activation = 'relu') \n",
    "    Dense_layer2 = Dense (f, activation = 'relu') \n",
    "    avg_out = Dense_layer2 (Dense_layer1(a_pool)) \n",
    "    max_out = Dense_layer2 (Dense_layer1( m_pool))\n",
    "    channel = add([ avg_out,max_out]) \n",
    "    channel = Activation('sigmoid') (channel) \n",
    "    channel = Reshape ((1,1,f)) (channel)  \n",
    "    C_out = Multiply()([x,channel])\n",
    "   \n",
    "    \n",
    "                      \n",
    "    ###### Spatial Attention ########\n",
    "    av_pooling=AveragePooling2D(pool_size=(1,1))(C_out)\n",
    "    ma_pooling=MaxPooling2D(pool_size=(1,1))(C_out)\n",
    "    spatial=concatenate([av_pooling,ma_pooling], axis = 3)\n",
    "    spatial =Conv2D(1,(3,3),strides=1,padding ='same') (spatial) \n",
    "    #spatial=ASPP(x,spatial)\n",
    "    \n",
    "    spatial= Activation('sigmoid')(spatial)\n",
    "      \n",
    "    S_out = Multiply()([C_out,spatial])\n",
    "    \n",
    "    return S_out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def conv_block(input, num_filters):\n",
    "    x = Conv2D(num_filters, 3, padding=\"same\")(input)\n",
    "    \n",
    "    \n",
    "    \n",
    "    x = Activation(\"relu\")(x)\n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_1 (InputLayer)           [(None, 8, 129, 6)]  0           []                               \n",
      "                                                                                                  \n",
      " conv2d (Conv2D)                (None, 8, 129, 16)   880         ['input_1[0][0]']                \n",
      "                                                                                                  \n",
      " activation (Activation)        (None, 8, 129, 16)   0           ['conv2d[0][0]']                 \n",
      "                                                                                                  \n",
      " max_pooling2d_1 (MaxPooling2D)  (None, 4, 64, 16)   0           ['activation[0][0]']             \n",
      "                                                                                                  \n",
      " conv2d_2 (Conv2D)              (None, 4, 64, 32)    4640        ['max_pooling2d_1[0][0]']        \n",
      "                                                                                                  \n",
      " activation_3 (Activation)      (None, 4, 64, 32)    0           ['conv2d_2[0][0]']               \n",
      "                                                                                                  \n",
      " dropout (Dropout)              (None, 4, 64, 32)    0           ['activation_3[0][0]']           \n",
      "                                                                                                  \n",
      " global_average_pooling2d_1 (Gl  (None, 32)          0           ['dropout[0][0]']                \n",
      " obalAveragePooling2D)                                                                            \n",
      "                                                                                                  \n",
      " global_max_pooling2d_1 (Global  (None, 32)          0           ['dropout[0][0]']                \n",
      " MaxPooling2D)                                                                                    \n",
      "                                                                                                  \n",
      " dense_2 (Dense)                (None, 4)            132         ['global_average_pooling2d_1[0][0\n",
      "                                                                 ]',                              \n",
      "                                                                  'global_max_pooling2d_1[0][0]'] \n",
      "                                                                                                  \n",
      " dense_3 (Dense)                (None, 32)           160         ['dense_2[0][0]',                \n",
      "                                                                  'dense_2[1][0]']                \n",
      "                                                                                                  \n",
      " add_1 (Add)                    (None, 32)           0           ['dense_3[0][0]',                \n",
      "                                                                  'dense_3[1][0]']                \n",
      "                                                                                                  \n",
      " activation_4 (Activation)      (None, 32)           0           ['add_1[0][0]']                  \n",
      "                                                                                                  \n",
      " reshape_1 (Reshape)            (None, 1, 1, 32)     0           ['activation_4[0][0]']           \n",
      "                                                                                                  \n",
      " multiply_2 (Multiply)          (None, 4, 64, 32)    0           ['dropout[0][0]',                \n",
      "                                                                  'reshape_1[0][0]']              \n",
      "                                                                                                  \n",
      " average_pooling2d_1 (AveragePo  (None, 4, 64, 32)   0           ['multiply_2[0][0]']             \n",
      " oling2D)                                                                                         \n",
      "                                                                                                  \n",
      " max_pooling2d_2 (MaxPooling2D)  (None, 4, 64, 32)   0           ['multiply_2[0][0]']             \n",
      "                                                                                                  \n",
      " concatenate_1 (Concatenate)    (None, 4, 64, 64)    0           ['average_pooling2d_1[0][0]',    \n",
      "                                                                  'max_pooling2d_2[0][0]']        \n",
      "                                                                                                  \n",
      " conv2d_3 (Conv2D)              (None, 4, 64, 1)     577         ['concatenate_1[0][0]']          \n",
      "                                                                                                  \n",
      " activation_5 (Activation)      (None, 4, 64, 1)     0           ['conv2d_3[0][0]']               \n",
      "                                                                                                  \n",
      " multiply_3 (Multiply)          (None, 4, 64, 32)    0           ['multiply_2[0][0]',             \n",
      "                                                                  'activation_5[0][0]']           \n",
      "                                                                                                  \n",
      " conv2d_4 (Conv2D)              (None, 4, 64, 32)    9248        ['multiply_3[0][0]']             \n",
      "                                                                                                  \n",
      " activation_6 (Activation)      (None, 4, 64, 32)    0           ['conv2d_4[0][0]']               \n",
      "                                                                                                  \n",
      " dropout_1 (Dropout)            (None, 4, 64, 32)    0           ['activation_6[0][0]']           \n",
      "                                                                                                  \n",
      " global_average_pooling2d_2 (Gl  (None, 32)          0           ['dropout_1[0][0]']              \n",
      " obalAveragePooling2D)                                                                            \n",
      "                                                                                                  \n",
      " global_max_pooling2d_2 (Global  (None, 32)          0           ['dropout_1[0][0]']              \n",
      " MaxPooling2D)                                                                                    \n",
      "                                                                                                  \n",
      " dense_4 (Dense)                (None, 4)            132         ['global_average_pooling2d_2[0][0\n",
      "                                                                 ]',                              \n",
      "                                                                  'global_max_pooling2d_2[0][0]'] \n",
      "                                                                                                  \n",
      " dense_5 (Dense)                (None, 32)           160         ['dense_4[0][0]',                \n",
      "                                                                  'dense_4[1][0]']                \n",
      "                                                                                                  \n",
      " add_2 (Add)                    (None, 32)           0           ['dense_5[0][0]',                \n",
      "                                                                  'dense_5[1][0]']                \n",
      "                                                                                                  \n",
      " activation_7 (Activation)      (None, 32)           0           ['add_2[0][0]']                  \n",
      "                                                                                                  \n",
      " reshape_2 (Reshape)            (None, 1, 1, 32)     0           ['activation_7[0][0]']           \n",
      "                                                                                                  \n",
      " multiply_4 (Multiply)          (None, 4, 64, 32)    0           ['dropout_1[0][0]',              \n",
      "                                                                  'reshape_2[0][0]']              \n",
      "                                                                                                  \n",
      " average_pooling2d_2 (AveragePo  (None, 4, 64, 32)   0           ['multiply_4[0][0]']             \n",
      " oling2D)                                                                                         \n",
      "                                                                                                  \n",
      " max_pooling2d_3 (MaxPooling2D)  (None, 4, 64, 32)   0           ['multiply_4[0][0]']             \n",
      "                                                                                                  \n",
      " concatenate_2 (Concatenate)    (None, 4, 64, 64)    0           ['average_pooling2d_2[0][0]',    \n",
      "                                                                  'max_pooling2d_3[0][0]']        \n",
      "                                                                                                  \n",
      " conv2d_5 (Conv2D)              (None, 4, 64, 1)     577         ['concatenate_2[0][0]']          \n",
      "                                                                                                  \n",
      " activation_8 (Activation)      (None, 4, 64, 1)     0           ['conv2d_5[0][0]']               \n",
      "                                                                                                  \n",
      " multiply_5 (Multiply)          (None, 4, 64, 32)    0           ['multiply_4[0][0]',             \n",
      "                                                                  'activation_8[0][0]']           \n",
      "                                                                                                  \n",
      " add_3 (Add)                    (None, 4, 64, 32)    0           ['multiply_3[0][0]',             \n",
      "                                                                  'multiply_5[0][0]']             \n",
      "                                                                                                  \n",
      " max_pooling2d_4 (MaxPooling2D)  (None, 2, 32, 32)   0           ['add_3[0][0]']                  \n",
      "                                                                                                  \n",
      " conv2d_6 (Conv2D)              (None, 2, 32, 64)    18496       ['max_pooling2d_4[0][0]']        \n",
      "                                                                                                  \n",
      " activation_9 (Activation)      (None, 2, 32, 64)    0           ['conv2d_6[0][0]']               \n",
      "                                                                                                  \n",
      " dropout_2 (Dropout)            (None, 2, 32, 64)    0           ['activation_9[0][0]']           \n",
      "                                                                                                  \n",
      " global_average_pooling2d_3 (Gl  (None, 64)          0           ['dropout_2[0][0]']              \n",
      " obalAveragePooling2D)                                                                            \n",
      "                                                                                                  \n",
      " global_max_pooling2d_3 (Global  (None, 64)          0           ['dropout_2[0][0]']              \n",
      " MaxPooling2D)                                                                                    \n",
      "                                                                                                  \n",
      " dense_6 (Dense)                (None, 8)            520         ['global_average_pooling2d_3[0][0\n",
      "                                                                 ]',                              \n",
      "                                                                  'global_max_pooling2d_3[0][0]'] \n",
      "                                                                                                  \n",
      " dense_7 (Dense)                (None, 64)           576         ['dense_6[0][0]',                \n",
      "                                                                  'dense_6[1][0]']                \n",
      "                                                                                                  \n",
      " add_4 (Add)                    (None, 64)           0           ['dense_7[0][0]',                \n",
      "                                                                  'dense_7[1][0]']                \n",
      "                                                                                                  \n",
      " activation_10 (Activation)     (None, 64)           0           ['add_4[0][0]']                  \n",
      "                                                                                                  \n",
      " reshape_3 (Reshape)            (None, 1, 1, 64)     0           ['activation_10[0][0]']          \n",
      "                                                                                                  \n",
      " multiply_6 (Multiply)          (None, 2, 32, 64)    0           ['dropout_2[0][0]',              \n",
      "                                                                  'reshape_3[0][0]']              \n",
      "                                                                                                  \n",
      " average_pooling2d_3 (AveragePo  (None, 2, 32, 64)   0           ['multiply_6[0][0]']             \n",
      " oling2D)                                                                                         \n",
      "                                                                                                  \n",
      " max_pooling2d_5 (MaxPooling2D)  (None, 2, 32, 64)   0           ['multiply_6[0][0]']             \n",
      "                                                                                                  \n",
      " concatenate_3 (Concatenate)    (None, 2, 32, 128)   0           ['average_pooling2d_3[0][0]',    \n",
      "                                                                  'max_pooling2d_5[0][0]']        \n",
      "                                                                                                  \n",
      " conv2d_7 (Conv2D)              (None, 2, 32, 1)     1153        ['concatenate_3[0][0]']          \n",
      "                                                                                                  \n",
      " activation_11 (Activation)     (None, 2, 32, 1)     0           ['conv2d_7[0][0]']               \n",
      "                                                                                                  \n",
      " multiply_7 (Multiply)          (None, 2, 32, 64)    0           ['multiply_6[0][0]',             \n",
      "                                                                  'activation_11[0][0]']          \n",
      "                                                                                                  \n",
      " conv2d_8 (Conv2D)              (None, 2, 32, 64)    36928       ['multiply_7[0][0]']             \n",
      "                                                                                                  \n",
      " activation_12 (Activation)     (None, 2, 32, 64)    0           ['conv2d_8[0][0]']               \n",
      "                                                                                                  \n",
      " dropout_3 (Dropout)            (None, 2, 32, 64)    0           ['activation_12[0][0]']          \n",
      "                                                                                                  \n",
      " global_average_pooling2d_4 (Gl  (None, 64)          0           ['dropout_3[0][0]']              \n",
      " obalAveragePooling2D)                                                                            \n",
      "                                                                                                  \n",
      " global_max_pooling2d_4 (Global  (None, 64)          0           ['dropout_3[0][0]']              \n",
      " MaxPooling2D)                                                                                    \n",
      "                                                                                                  \n",
      " dense_8 (Dense)                (None, 8)            520         ['global_average_pooling2d_4[0][0\n",
      "                                                                 ]',                              \n",
      "                                                                  'global_max_pooling2d_4[0][0]'] \n",
      "                                                                                                  \n",
      " dense_9 (Dense)                (None, 64)           576         ['dense_8[0][0]',                \n",
      "                                                                  'dense_8[1][0]']                \n",
      "                                                                                                  \n",
      " add_5 (Add)                    (None, 64)           0           ['dense_9[0][0]',                \n",
      "                                                                  'dense_9[1][0]']                \n",
      "                                                                                                  \n",
      " activation_13 (Activation)     (None, 64)           0           ['add_5[0][0]']                  \n",
      "                                                                                                  \n",
      " reshape_4 (Reshape)            (None, 1, 1, 64)     0           ['activation_13[0][0]']          \n",
      "                                                                                                  \n",
      " multiply_8 (Multiply)          (None, 2, 32, 64)    0           ['dropout_3[0][0]',              \n",
      "                                                                  'reshape_4[0][0]']              \n",
      "                                                                                                  \n",
      " average_pooling2d_4 (AveragePo  (None, 2, 32, 64)   0           ['multiply_8[0][0]']             \n",
      " oling2D)                                                                                         \n",
      "                                                                                                  \n",
      " max_pooling2d_6 (MaxPooling2D)  (None, 2, 32, 64)   0           ['multiply_8[0][0]']             \n",
      "                                                                                                  \n",
      " concatenate_4 (Concatenate)    (None, 2, 32, 128)   0           ['average_pooling2d_4[0][0]',    \n",
      "                                                                  'max_pooling2d_6[0][0]']        \n",
      "                                                                                                  \n",
      " conv2d_9 (Conv2D)              (None, 2, 32, 1)     1153        ['concatenate_4[0][0]']          \n",
      "                                                                                                  \n",
      " activation_14 (Activation)     (None, 2, 32, 1)     0           ['conv2d_9[0][0]']               \n",
      "                                                                                                  \n",
      " multiply_9 (Multiply)          (None, 2, 32, 64)    0           ['multiply_8[0][0]',             \n",
      "                                                                  'activation_14[0][0]']          \n",
      "                                                                                                  \n",
      " add_6 (Add)                    (None, 2, 32, 64)    0           ['multiply_7[0][0]',             \n",
      "                                                                  'multiply_9[0][0]']             \n",
      "                                                                                                  \n",
      " max_pooling2d_7 (MaxPooling2D)  (None, 1, 16, 64)   0           ['add_6[0][0]']                  \n",
      "                                                                                                  \n",
      " conv2d_10 (Conv2D)             (None, 1, 16, 128)   73856       ['max_pooling2d_7[0][0]']        \n",
      "                                                                                                  \n",
      " activation_15 (Activation)     (None, 1, 16, 128)   0           ['conv2d_10[0][0]']              \n",
      "                                                                                                  \n",
      " dropout_4 (Dropout)            (None, 1, 16, 128)   0           ['activation_15[0][0]']          \n",
      "                                                                                                  \n",
      " flatten (Flatten)              (None, 2048)         0           ['dropout_4[0][0]']              \n",
      "                                                                                                  \n",
      " dense_12 (Dense)               (None, 48)           98352       ['flatten[0][0]']                \n",
      "                                                                                                  \n",
      " dropout_7 (Dropout)            (None, 48)           0           ['dense_12[0][0]']               \n",
      "                                                                                                  \n",
      " dense_13 (Dense)               (None, 32)           1568        ['dropout_7[0][0]']              \n",
      "                                                                                                  \n",
      " dense_14 (Dense)               (None, 18)           594         ['dense_13[0][0]']               \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 250,798\n",
      "Trainable params: 250,798\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Khan\\anaconda3\\envs\\gputest\\lib\\site-packages\\keras\\optimizers\\optimizer_v2\\adam.py:110: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
      "  super(Adam, self).__init__(name, **kwargs)\n"
     ]
    }
   ],
   "source": [
    "inputA = Input(shape=X_train[0].shape)\n",
    "\n",
    "\n",
    "##### Channel 1 #########\n",
    "s1 = conv_block(inputA,16)\n",
    "a1=CBAM(s1,8)\n",
    "pool1 = MaxPooling2D(pool_size=(2, 2))(s1)\n",
    "\n",
    "s2 = conv_block(pool1,32)\n",
    "s2=Dropout(0.1)(s2)\n",
    "a2=CBAM(s2,8)\n",
    "\n",
    "\n",
    "\n",
    "#pool2 = MaxPooling2D(pool_size=(2, 2))(a2)\n",
    "\n",
    "s3 = conv_block(a2, 32)\n",
    "s3=Dropout(0.1)(s3)\n",
    "a3=CBAM(s3,8)\n",
    "\n",
    "\n",
    "a33=Add()([a2,a3])\n",
    "pool3 = MaxPooling2D(pool_size=(2, 2))(a33)\n",
    "\n",
    "s4= conv_block(pool3,64)\n",
    "s4=Dropout(0.15)(s4)\n",
    "a4=CBAM(s4,8)\n",
    "\n",
    "s5= conv_block(a4,64)\n",
    "s5=Dropout(0.15)(s5)\n",
    "a5=CBAM(s5,8)\n",
    "\n",
    "a55=Add()([a4,a5])\n",
    "\n",
    "\n",
    "\n",
    "pool5 = MaxPooling2D(pool_size=(2,2))(a55)\n",
    "\n",
    "s6= conv_block(pool5,128)\n",
    "\n",
    "s7=Dropout(0.15)(s6)\n",
    "\n",
    "###### Concatenation ##############\n",
    "\n",
    "F=Flatten()(s7)\n",
    "\n",
    "\n",
    "x = Dense(128, activation=\"relu\")(F)\n",
    "x=Dropout(0.15)(x)\n",
    "\n",
    "x = Dense(64, activation=\"relu\")(F)\n",
    "x=Dropout(0.15)(x)\n",
    "\n",
    "x = Dense(48, activation=\"relu\")(F)\n",
    "x=Dropout(0.1)(x)\n",
    "\n",
    "x = Dense(32, activation=\"relu\")(x)\n",
    "\n",
    "\n",
    "\n",
    "O = Dense(18, activation=\"softmax\")(x)\n",
    "\n",
    "model =  keras.Model(inputs=inputA, outputs=O)\n",
    "opt = keras.optimizers.Adam(lr=1e-4)\n",
    "model.compile(optimizer=opt,  loss = 'mse', metrics = ['accuracy','precision_m', 'recall_m', 'f1_m'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.\n"
     ]
    }
   ],
   "source": [
    "model.save('D:/Projects/Anni apu/Signal.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import load_model\n",
    " \n",
    "# load model\n",
    "model = load_model('D:/Projects/Anni apu/Signal.h5',custom_objects={'precision_m':precision_m, 'recall_m':recall_m, 'f1_m':f1_m})\n",
    "# summarize model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "487/487 [==============================] - 10s 11ms/step - loss: 0.0507 - accuracy: 0.2288 - precision_m: 0.3128 - recall_m: 0.0171 - f1_m: 0.0319 - val_loss: 0.0458 - val_accuracy: 0.3242 - val_precision_m: 0.5413 - val_recall_m: 0.0449 - val_f1_m: 0.0814\n",
      "Epoch 2/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0404 - accuracy: 0.4071 - precision_m: 0.6820 - recall_m: 0.1630 - f1_m: 0.2553 - val_loss: 0.0348 - val_accuracy: 0.4975 - val_precision_m: 0.8051 - val_recall_m: 0.2398 - val_f1_m: 0.3642\n",
      "Epoch 3/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0335 - accuracy: 0.5069 - precision_m: 0.7841 - recall_m: 0.3074 - f1_m: 0.4370 - val_loss: 0.0293 - val_accuracy: 0.6170 - val_precision_m: 0.8913 - val_recall_m: 0.3587 - val_f1_m: 0.5067\n",
      "Epoch 4/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0294 - accuracy: 0.5809 - precision_m: 0.8255 - recall_m: 0.3900 - f1_m: 0.5259 - val_loss: 0.0272 - val_accuracy: 0.6276 - val_precision_m: 0.8581 - val_recall_m: 0.3992 - val_f1_m: 0.5402\n",
      "Epoch 5/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0263 - accuracy: 0.6254 - precision_m: 0.8494 - recall_m: 0.4586 - f1_m: 0.5924 - val_loss: 0.0234 - val_accuracy: 0.6769 - val_precision_m: 0.9126 - val_recall_m: 0.4676 - val_f1_m: 0.6137\n",
      "Epoch 6/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0237 - accuracy: 0.6595 - precision_m: 0.8653 - recall_m: 0.5096 - f1_m: 0.6378 - val_loss: 0.0214 - val_accuracy: 0.7120 - val_precision_m: 0.9091 - val_recall_m: 0.5272 - val_f1_m: 0.6635\n",
      "Epoch 7/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0225 - accuracy: 0.6828 - precision_m: 0.8676 - recall_m: 0.5372 - f1_m: 0.6603 - val_loss: 0.0202 - val_accuracy: 0.7203 - val_precision_m: 0.9101 - val_recall_m: 0.5489 - val_f1_m: 0.6815\n",
      "Epoch 8/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0209 - accuracy: 0.7092 - precision_m: 0.8782 - recall_m: 0.5711 - f1_m: 0.6888 - val_loss: 0.0192 - val_accuracy: 0.7392 - val_precision_m: 0.9112 - val_recall_m: 0.5713 - val_f1_m: 0.6992\n",
      "Epoch 9/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0196 - accuracy: 0.7326 - precision_m: 0.8785 - recall_m: 0.6085 - f1_m: 0.7164 - val_loss: 0.0182 - val_accuracy: 0.7594 - val_precision_m: 0.9104 - val_recall_m: 0.6062 - val_f1_m: 0.7248\n",
      "Epoch 10/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0188 - accuracy: 0.7464 - precision_m: 0.8748 - recall_m: 0.6313 - f1_m: 0.7310 - val_loss: 0.0177 - val_accuracy: 0.7733 - val_precision_m: 0.9042 - val_recall_m: 0.6382 - val_f1_m: 0.7456\n",
      "Epoch 11/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0180 - accuracy: 0.7637 - precision_m: 0.8753 - recall_m: 0.6572 - f1_m: 0.7488 - val_loss: 0.0168 - val_accuracy: 0.7751 - val_precision_m: 0.8781 - val_recall_m: 0.6896 - val_f1_m: 0.7705\n",
      "Epoch 12/200\n",
      "487/487 [==============================] - 6s 12ms/step - loss: 0.0177 - accuracy: 0.7669 - precision_m: 0.8656 - recall_m: 0.6746 - f1_m: 0.7565 - val_loss: 0.0160 - val_accuracy: 0.8028 - val_precision_m: 0.9053 - val_recall_m: 0.6697 - val_f1_m: 0.7678\n",
      "Epoch 13/200\n",
      "487/487 [==============================] - 6s 11ms/step - loss: 0.0167 - accuracy: 0.7845 - precision_m: 0.8711 - recall_m: 0.6945 - f1_m: 0.7712 - val_loss: 0.0156 - val_accuracy: 0.8011 - val_precision_m: 0.8861 - val_recall_m: 0.7030 - val_f1_m: 0.7823\n",
      "Epoch 14/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0160 - accuracy: 0.7906 - precision_m: 0.8739 - recall_m: 0.7143 - f1_m: 0.7847 - val_loss: 0.0151 - val_accuracy: 0.8076 - val_precision_m: 0.8978 - val_recall_m: 0.7111 - val_f1_m: 0.7917\n",
      "Epoch 15/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0153 - accuracy: 0.8056 - precision_m: 0.8757 - recall_m: 0.7336 - f1_m: 0.7971 - val_loss: 0.0145 - val_accuracy: 0.8190 - val_precision_m: 0.9024 - val_recall_m: 0.7360 - val_f1_m: 0.8093\n",
      "Epoch 16/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0146 - accuracy: 0.8165 - precision_m: 0.8804 - recall_m: 0.7545 - f1_m: 0.8116 - val_loss: 0.0138 - val_accuracy: 0.8356 - val_precision_m: 0.8952 - val_recall_m: 0.7569 - val_f1_m: 0.8191\n",
      "Epoch 17/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0141 - accuracy: 0.8255 - precision_m: 0.8775 - recall_m: 0.7736 - f1_m: 0.8213 - val_loss: 0.0134 - val_accuracy: 0.8398 - val_precision_m: 0.8980 - val_recall_m: 0.7738 - val_f1_m: 0.8298\n",
      "Epoch 18/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0134 - accuracy: 0.8375 - precision_m: 0.8819 - recall_m: 0.7928 - f1_m: 0.8342 - val_loss: 0.0124 - val_accuracy: 0.8520 - val_precision_m: 0.8971 - val_recall_m: 0.8081 - val_f1_m: 0.8495\n",
      "Epoch 19/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0125 - accuracy: 0.8461 - precision_m: 0.8841 - recall_m: 0.8107 - f1_m: 0.8452 - val_loss: 0.0125 - val_accuracy: 0.8479 - val_precision_m: 0.8818 - val_recall_m: 0.8165 - val_f1_m: 0.8473\n",
      "Epoch 20/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0119 - accuracy: 0.8588 - precision_m: 0.8924 - recall_m: 0.8238 - f1_m: 0.8562 - val_loss: 0.0112 - val_accuracy: 0.8653 - val_precision_m: 0.9010 - val_recall_m: 0.8311 - val_f1_m: 0.8640\n",
      "Epoch 21/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0114 - accuracy: 0.8640 - precision_m: 0.8932 - recall_m: 0.8367 - f1_m: 0.8635 - val_loss: 0.0108 - val_accuracy: 0.8707 - val_precision_m: 0.9079 - val_recall_m: 0.8367 - val_f1_m: 0.8701\n",
      "Epoch 22/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0108 - accuracy: 0.8685 - precision_m: 0.8974 - recall_m: 0.8437 - f1_m: 0.8693 - val_loss: 0.0100 - val_accuracy: 0.8836 - val_precision_m: 0.9126 - val_recall_m: 0.8526 - val_f1_m: 0.8810\n",
      "Epoch 23/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0103 - accuracy: 0.8760 - precision_m: 0.9024 - recall_m: 0.8505 - f1_m: 0.8752 - val_loss: 0.0097 - val_accuracy: 0.8892 - val_precision_m: 0.9183 - val_recall_m: 0.8574 - val_f1_m: 0.8862\n",
      "Epoch 24/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0099 - accuracy: 0.8821 - precision_m: 0.9060 - recall_m: 0.8595 - f1_m: 0.8817 - val_loss: 0.0093 - val_accuracy: 0.8921 - val_precision_m: 0.9185 - val_recall_m: 0.8635 - val_f1_m: 0.8896\n",
      "Epoch 25/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0093 - accuracy: 0.8886 - precision_m: 0.9098 - recall_m: 0.8701 - f1_m: 0.8892 - val_loss: 0.0087 - val_accuracy: 0.9046 - val_precision_m: 0.9304 - val_recall_m: 0.8760 - val_f1_m: 0.9019\n",
      "Epoch 26/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0091 - accuracy: 0.8954 - precision_m: 0.9142 - recall_m: 0.8742 - f1_m: 0.8934 - val_loss: 0.0083 - val_accuracy: 0.9050 - val_precision_m: 0.9248 - val_recall_m: 0.8802 - val_f1_m: 0.9016\n",
      "Epoch 27/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0087 - accuracy: 0.8976 - precision_m: 0.9182 - recall_m: 0.8785 - f1_m: 0.8975 - val_loss: 0.0081 - val_accuracy: 0.9098 - val_precision_m: 0.9313 - val_recall_m: 0.8900 - val_f1_m: 0.9098\n",
      "Epoch 28/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0085 - accuracy: 0.8990 - precision_m: 0.9175 - recall_m: 0.8836 - f1_m: 0.9000 - val_loss: 0.0080 - val_accuracy: 0.9100 - val_precision_m: 0.9279 - val_recall_m: 0.8913 - val_f1_m: 0.9089\n",
      "Epoch 29/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0082 - accuracy: 0.9048 - precision_m: 0.9224 - recall_m: 0.8888 - f1_m: 0.9050 - val_loss: 0.0083 - val_accuracy: 0.9015 - val_precision_m: 0.9184 - val_recall_m: 0.8832 - val_f1_m: 0.9002\n",
      "Epoch 30/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0078 - accuracy: 0.9104 - precision_m: 0.9250 - recall_m: 0.8954 - f1_m: 0.9097 - val_loss: 0.0075 - val_accuracy: 0.9146 - val_precision_m: 0.9320 - val_recall_m: 0.8982 - val_f1_m: 0.9145\n",
      "Epoch 31/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0076 - accuracy: 0.9118 - precision_m: 0.9293 - recall_m: 0.8987 - f1_m: 0.9135 - val_loss: 0.0073 - val_accuracy: 0.9165 - val_precision_m: 0.9365 - val_recall_m: 0.9013 - val_f1_m: 0.9183\n",
      "Epoch 32/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0072 - accuracy: 0.9158 - precision_m: 0.9311 - recall_m: 0.9021 - f1_m: 0.9161 - val_loss: 0.0071 - val_accuracy: 0.9189 - val_precision_m: 0.9336 - val_recall_m: 0.9072 - val_f1_m: 0.9200\n",
      "Epoch 33/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0072 - accuracy: 0.9171 - precision_m: 0.9319 - recall_m: 0.9046 - f1_m: 0.9178 - val_loss: 0.0070 - val_accuracy: 0.9223 - val_precision_m: 0.9357 - val_recall_m: 0.9086 - val_f1_m: 0.9217\n",
      "Epoch 34/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0069 - accuracy: 0.9198 - precision_m: 0.9354 - recall_m: 0.9075 - f1_m: 0.9210 - val_loss: 0.0075 - val_accuracy: 0.9117 - val_precision_m: 0.9288 - val_recall_m: 0.8986 - val_f1_m: 0.9132\n",
      "Epoch 35/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0067 - accuracy: 0.9231 - precision_m: 0.9364 - recall_m: 0.9112 - f1_m: 0.9234 - val_loss: 0.0068 - val_accuracy: 0.9233 - val_precision_m: 0.9402 - val_recall_m: 0.9078 - val_f1_m: 0.9235\n",
      "Epoch 36/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0068 - accuracy: 0.9221 - precision_m: 0.9350 - recall_m: 0.9104 - f1_m: 0.9223 - val_loss: 0.0071 - val_accuracy: 0.9200 - val_precision_m: 0.9323 - val_recall_m: 0.9057 - val_f1_m: 0.9185\n",
      "Epoch 37/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0065 - accuracy: 0.9256 - precision_m: 0.9389 - recall_m: 0.9157 - f1_m: 0.9269 - val_loss: 0.0070 - val_accuracy: 0.9202 - val_precision_m: 0.9311 - val_recall_m: 0.9099 - val_f1_m: 0.9202\n",
      "Epoch 38/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0064 - accuracy: 0.9257 - precision_m: 0.9382 - recall_m: 0.9153 - f1_m: 0.9264 - val_loss: 0.0065 - val_accuracy: 0.9285 - val_precision_m: 0.9380 - val_recall_m: 0.9155 - val_f1_m: 0.9264\n",
      "Epoch 39/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0061 - accuracy: 0.9293 - precision_m: 0.9413 - recall_m: 0.9204 - f1_m: 0.9306 - val_loss: 0.0062 - val_accuracy: 0.9295 - val_precision_m: 0.9423 - val_recall_m: 0.9170 - val_f1_m: 0.9292\n",
      "Epoch 40/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0061 - accuracy: 0.9319 - precision_m: 0.9435 - recall_m: 0.9206 - f1_m: 0.9318 - val_loss: 0.0058 - val_accuracy: 0.9341 - val_precision_m: 0.9471 - val_recall_m: 0.9231 - val_f1_m: 0.9348\n",
      "Epoch 41/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0058 - accuracy: 0.9332 - precision_m: 0.9445 - recall_m: 0.9232 - f1_m: 0.9336 - val_loss: 0.0062 - val_accuracy: 0.9287 - val_precision_m: 0.9404 - val_recall_m: 0.9206 - val_f1_m: 0.9303\n",
      "Epoch 42/200\n",
      "487/487 [==============================] - 6s 12ms/step - loss: 0.0058 - accuracy: 0.9343 - precision_m: 0.9441 - recall_m: 0.9247 - f1_m: 0.9341 - val_loss: 0.0060 - val_accuracy: 0.9320 - val_precision_m: 0.9440 - val_recall_m: 0.9245 - val_f1_m: 0.9340\n",
      "Epoch 43/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0055 - accuracy: 0.9382 - precision_m: 0.9480 - recall_m: 0.9302 - f1_m: 0.9389 - val_loss: 0.0056 - val_accuracy: 0.9375 - val_precision_m: 0.9480 - val_recall_m: 0.9256 - val_f1_m: 0.9365\n",
      "Epoch 44/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0054 - accuracy: 0.9382 - precision_m: 0.9488 - recall_m: 0.9294 - f1_m: 0.9388 - val_loss: 0.0058 - val_accuracy: 0.9329 - val_precision_m: 0.9457 - val_recall_m: 0.9243 - val_f1_m: 0.9347\n",
      "Epoch 45/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0052 - accuracy: 0.9404 - precision_m: 0.9493 - recall_m: 0.9318 - f1_m: 0.9403 - val_loss: 0.0054 - val_accuracy: 0.9377 - val_precision_m: 0.9454 - val_recall_m: 0.9304 - val_f1_m: 0.9377\n",
      "Epoch 46/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0052 - accuracy: 0.9404 - precision_m: 0.9502 - recall_m: 0.9322 - f1_m: 0.9410 - val_loss: 0.0058 - val_accuracy: 0.9327 - val_precision_m: 0.9415 - val_recall_m: 0.9248 - val_f1_m: 0.9330\n",
      "Epoch 47/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0050 - accuracy: 0.9433 - precision_m: 0.9519 - recall_m: 0.9354 - f1_m: 0.9435 - val_loss: 0.0054 - val_accuracy: 0.9391 - val_precision_m: 0.9487 - val_recall_m: 0.9310 - val_f1_m: 0.9396\n",
      "Epoch 48/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0050 - accuracy: 0.9442 - precision_m: 0.9518 - recall_m: 0.9367 - f1_m: 0.9441 - val_loss: 0.0053 - val_accuracy: 0.9404 - val_precision_m: 0.9519 - val_recall_m: 0.9321 - val_f1_m: 0.9418\n",
      "Epoch 49/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0049 - accuracy: 0.9453 - precision_m: 0.9537 - recall_m: 0.9379 - f1_m: 0.9456 - val_loss: 0.0053 - val_accuracy: 0.9416 - val_precision_m: 0.9489 - val_recall_m: 0.9337 - val_f1_m: 0.9411\n",
      "Epoch 50/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0049 - accuracy: 0.9431 - precision_m: 0.9505 - recall_m: 0.9356 - f1_m: 0.9429 - val_loss: 0.0055 - val_accuracy: 0.9397 - val_precision_m: 0.9476 - val_recall_m: 0.9306 - val_f1_m: 0.9389\n",
      "Epoch 51/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0046 - accuracy: 0.9485 - precision_m: 0.9563 - recall_m: 0.9411 - f1_m: 0.9485 - val_loss: 0.0054 - val_accuracy: 0.9404 - val_precision_m: 0.9462 - val_recall_m: 0.9333 - val_f1_m: 0.9396\n",
      "Epoch 52/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0046 - accuracy: 0.9492 - precision_m: 0.9555 - recall_m: 0.9408 - f1_m: 0.9480 - val_loss: 0.0050 - val_accuracy: 0.9447 - val_precision_m: 0.9537 - val_recall_m: 0.9363 - val_f1_m: 0.9448\n",
      "Epoch 53/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0046 - accuracy: 0.9496 - precision_m: 0.9570 - recall_m: 0.9432 - f1_m: 0.9499 - val_loss: 0.0049 - val_accuracy: 0.9464 - val_precision_m: 0.9551 - val_recall_m: 0.9362 - val_f1_m: 0.9454\n",
      "Epoch 54/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0043 - accuracy: 0.9518 - precision_m: 0.9603 - recall_m: 0.9455 - f1_m: 0.9527 - val_loss: 0.0050 - val_accuracy: 0.9458 - val_precision_m: 0.9515 - val_recall_m: 0.9365 - val_f1_m: 0.9439\n",
      "Epoch 55/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0043 - accuracy: 0.9523 - precision_m: 0.9596 - recall_m: 0.9469 - f1_m: 0.9531 - val_loss: 0.0049 - val_accuracy: 0.9472 - val_precision_m: 0.9527 - val_recall_m: 0.9400 - val_f1_m: 0.9462\n",
      "Epoch 56/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0043 - accuracy: 0.9511 - precision_m: 0.9580 - recall_m: 0.9467 - f1_m: 0.9522 - val_loss: 0.0050 - val_accuracy: 0.9447 - val_precision_m: 0.9534 - val_recall_m: 0.9352 - val_f1_m: 0.9441\n",
      "Epoch 57/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0042 - accuracy: 0.9539 - precision_m: 0.9609 - recall_m: 0.9480 - f1_m: 0.9544 - val_loss: 0.0049 - val_accuracy: 0.9451 - val_precision_m: 0.9525 - val_recall_m: 0.9379 - val_f1_m: 0.9450\n",
      "Epoch 58/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0041 - accuracy: 0.9551 - precision_m: 0.9608 - recall_m: 0.9496 - f1_m: 0.9551 - val_loss: 0.0049 - val_accuracy: 0.9464 - val_precision_m: 0.9532 - val_recall_m: 0.9387 - val_f1_m: 0.9458\n",
      "Epoch 59/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0040 - accuracy: 0.9570 - precision_m: 0.9620 - recall_m: 0.9505 - f1_m: 0.9561 - val_loss: 0.0049 - val_accuracy: 0.9445 - val_precision_m: 0.9511 - val_recall_m: 0.9360 - val_f1_m: 0.9434\n",
      "Epoch 60/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0039 - accuracy: 0.9566 - precision_m: 0.9629 - recall_m: 0.9514 - f1_m: 0.9570 - val_loss: 0.0052 - val_accuracy: 0.9410 - val_precision_m: 0.9485 - val_recall_m: 0.9367 - val_f1_m: 0.9425\n",
      "Epoch 61/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0039 - accuracy: 0.9580 - precision_m: 0.9626 - recall_m: 0.9525 - f1_m: 0.9574 - val_loss: 0.0046 - val_accuracy: 0.9503 - val_precision_m: 0.9550 - val_recall_m: 0.9434 - val_f1_m: 0.9491\n",
      "Epoch 62/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0039 - accuracy: 0.9571 - precision_m: 0.9628 - recall_m: 0.9524 - f1_m: 0.9575 - val_loss: 0.0046 - val_accuracy: 0.9472 - val_precision_m: 0.9528 - val_recall_m: 0.9415 - val_f1_m: 0.9471\n",
      "Epoch 63/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0040 - accuracy: 0.9561 - precision_m: 0.9618 - recall_m: 0.9516 - f1_m: 0.9566 - val_loss: 0.0049 - val_accuracy: 0.9451 - val_precision_m: 0.9515 - val_recall_m: 0.9390 - val_f1_m: 0.9451\n",
      "Epoch 64/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0037 - accuracy: 0.9591 - precision_m: 0.9648 - recall_m: 0.9541 - f1_m: 0.9593 - val_loss: 0.0047 - val_accuracy: 0.9481 - val_precision_m: 0.9564 - val_recall_m: 0.9421 - val_f1_m: 0.9491\n",
      "Epoch 65/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0037 - accuracy: 0.9611 - precision_m: 0.9653 - recall_m: 0.9560 - f1_m: 0.9605 - val_loss: 0.0044 - val_accuracy: 0.9516 - val_precision_m: 0.9571 - val_recall_m: 0.9465 - val_f1_m: 0.9517\n",
      "Epoch 66/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0036 - accuracy: 0.9613 - precision_m: 0.9659 - recall_m: 0.9562 - f1_m: 0.9610 - val_loss: 0.0046 - val_accuracy: 0.9510 - val_precision_m: 0.9555 - val_recall_m: 0.9461 - val_f1_m: 0.9507\n",
      "Epoch 67/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0036 - accuracy: 0.9607 - precision_m: 0.9661 - recall_m: 0.9554 - f1_m: 0.9606 - val_loss: 0.0046 - val_accuracy: 0.9487 - val_precision_m: 0.9532 - val_recall_m: 0.9444 - val_f1_m: 0.9487\n",
      "Epoch 68/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0036 - accuracy: 0.9600 - precision_m: 0.9647 - recall_m: 0.9569 - f1_m: 0.9607 - val_loss: 0.0043 - val_accuracy: 0.9508 - val_precision_m: 0.9567 - val_recall_m: 0.9473 - val_f1_m: 0.9519\n",
      "Epoch 69/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0036 - accuracy: 0.9602 - precision_m: 0.9656 - recall_m: 0.9557 - f1_m: 0.9606 - val_loss: 0.0043 - val_accuracy: 0.9535 - val_precision_m: 0.9584 - val_recall_m: 0.9496 - val_f1_m: 0.9539\n",
      "Epoch 70/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0033 - accuracy: 0.9634 - precision_m: 0.9687 - recall_m: 0.9598 - f1_m: 0.9641 - val_loss: 0.0046 - val_accuracy: 0.9505 - val_precision_m: 0.9532 - val_recall_m: 0.9448 - val_f1_m: 0.9489\n",
      "Epoch 71/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0034 - accuracy: 0.9631 - precision_m: 0.9677 - recall_m: 0.9584 - f1_m: 0.9630 - val_loss: 0.0045 - val_accuracy: 0.9512 - val_precision_m: 0.9565 - val_recall_m: 0.9473 - val_f1_m: 0.9518\n",
      "Epoch 72/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0034 - accuracy: 0.9621 - precision_m: 0.9663 - recall_m: 0.9582 - f1_m: 0.9621 - val_loss: 0.0045 - val_accuracy: 0.9499 - val_precision_m: 0.9569 - val_recall_m: 0.9436 - val_f1_m: 0.9501\n",
      "Epoch 73/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0035 - accuracy: 0.9617 - precision_m: 0.9675 - recall_m: 0.9577 - f1_m: 0.9625 - val_loss: 0.0041 - val_accuracy: 0.9562 - val_precision_m: 0.9599 - val_recall_m: 0.9513 - val_f1_m: 0.9555\n",
      "Epoch 74/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0032 - accuracy: 0.9664 - precision_m: 0.9697 - recall_m: 0.9633 - f1_m: 0.9664 - val_loss: 0.0044 - val_accuracy: 0.9528 - val_precision_m: 0.9564 - val_recall_m: 0.9482 - val_f1_m: 0.9523\n",
      "Epoch 75/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0033 - accuracy: 0.9636 - precision_m: 0.9673 - recall_m: 0.9597 - f1_m: 0.9634 - val_loss: 0.0042 - val_accuracy: 0.9549 - val_precision_m: 0.9599 - val_recall_m: 0.9496 - val_f1_m: 0.9546\n",
      "Epoch 76/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0033 - accuracy: 0.9648 - precision_m: 0.9682 - recall_m: 0.9613 - f1_m: 0.9647 - val_loss: 0.0046 - val_accuracy: 0.9481 - val_precision_m: 0.9526 - val_recall_m: 0.9442 - val_f1_m: 0.9483\n",
      "Epoch 77/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0033 - accuracy: 0.9639 - precision_m: 0.9679 - recall_m: 0.9601 - f1_m: 0.9639 - val_loss: 0.0043 - val_accuracy: 0.9518 - val_precision_m: 0.9557 - val_recall_m: 0.9482 - val_f1_m: 0.9519\n",
      "Epoch 78/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0032 - accuracy: 0.9665 - precision_m: 0.9704 - recall_m: 0.9621 - f1_m: 0.9661 - val_loss: 0.0044 - val_accuracy: 0.9510 - val_precision_m: 0.9555 - val_recall_m: 0.9486 - val_f1_m: 0.9520\n",
      "Epoch 79/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0031 - accuracy: 0.9670 - precision_m: 0.9712 - recall_m: 0.9636 - f1_m: 0.9673 - val_loss: 0.0044 - val_accuracy: 0.9507 - val_precision_m: 0.9553 - val_recall_m: 0.9467 - val_f1_m: 0.9509\n",
      "Epoch 80/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0030 - accuracy: 0.9679 - precision_m: 0.9712 - recall_m: 0.9647 - f1_m: 0.9679 - val_loss: 0.0043 - val_accuracy: 0.9518 - val_precision_m: 0.9558 - val_recall_m: 0.9477 - val_f1_m: 0.9516\n",
      "Epoch 81/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0030 - accuracy: 0.9681 - precision_m: 0.9713 - recall_m: 0.9646 - f1_m: 0.9678 - val_loss: 0.0041 - val_accuracy: 0.9539 - val_precision_m: 0.9583 - val_recall_m: 0.9517 - val_f1_m: 0.9549\n",
      "Epoch 82/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0028 - accuracy: 0.9690 - precision_m: 0.9725 - recall_m: 0.9662 - f1_m: 0.9693 - val_loss: 0.0040 - val_accuracy: 0.9564 - val_precision_m: 0.9604 - val_recall_m: 0.9534 - val_f1_m: 0.9569\n",
      "Epoch 83/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0029 - accuracy: 0.9681 - precision_m: 0.9725 - recall_m: 0.9641 - f1_m: 0.9682 - val_loss: 0.0038 - val_accuracy: 0.9584 - val_precision_m: 0.9623 - val_recall_m: 0.9551 - val_f1_m: 0.9586\n",
      "Epoch 84/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0029 - accuracy: 0.9677 - precision_m: 0.9717 - recall_m: 0.9652 - f1_m: 0.9684 - val_loss: 0.0038 - val_accuracy: 0.9566 - val_precision_m: 0.9624 - val_recall_m: 0.9509 - val_f1_m: 0.9565\n",
      "Epoch 85/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0029 - accuracy: 0.9700 - precision_m: 0.9723 - recall_m: 0.9667 - f1_m: 0.9695 - val_loss: 0.0043 - val_accuracy: 0.9518 - val_precision_m: 0.9568 - val_recall_m: 0.9479 - val_f1_m: 0.9523\n",
      "Epoch 86/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0029 - accuracy: 0.9689 - precision_m: 0.9718 - recall_m: 0.9659 - f1_m: 0.9688 - val_loss: 0.0040 - val_accuracy: 0.9568 - val_precision_m: 0.9602 - val_recall_m: 0.9521 - val_f1_m: 0.9560\n",
      "Epoch 87/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0028 - accuracy: 0.9704 - precision_m: 0.9738 - recall_m: 0.9676 - f1_m: 0.9707 - val_loss: 0.0040 - val_accuracy: 0.9559 - val_precision_m: 0.9603 - val_recall_m: 0.9521 - val_f1_m: 0.9561\n",
      "Epoch 88/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0028 - accuracy: 0.9695 - precision_m: 0.9728 - recall_m: 0.9669 - f1_m: 0.9698 - val_loss: 0.0040 - val_accuracy: 0.9555 - val_precision_m: 0.9599 - val_recall_m: 0.9538 - val_f1_m: 0.9568\n",
      "Epoch 89/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0028 - accuracy: 0.9710 - precision_m: 0.9740 - recall_m: 0.9690 - f1_m: 0.9715 - val_loss: 0.0043 - val_accuracy: 0.9535 - val_precision_m: 0.9563 - val_recall_m: 0.9494 - val_f1_m: 0.9528\n",
      "Epoch 90/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0026 - accuracy: 0.9725 - precision_m: 0.9754 - recall_m: 0.9696 - f1_m: 0.9725 - val_loss: 0.0041 - val_accuracy: 0.9520 - val_precision_m: 0.9566 - val_recall_m: 0.9482 - val_f1_m: 0.9524\n",
      "Epoch 91/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0037 - accuracy: 0.9595 - precision_m: 0.9624 - recall_m: 0.9559 - f1_m: 0.9591 - val_loss: 0.0041 - val_accuracy: 0.9562 - val_precision_m: 0.9602 - val_recall_m: 0.9521 - val_f1_m: 0.9560\n",
      "Epoch 92/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0029 - accuracy: 0.9679 - precision_m: 0.9716 - recall_m: 0.9652 - f1_m: 0.9683 - val_loss: 0.0039 - val_accuracy: 0.9582 - val_precision_m: 0.9610 - val_recall_m: 0.9553 - val_f1_m: 0.9581\n",
      "Epoch 93/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0025 - accuracy: 0.9736 - precision_m: 0.9759 - recall_m: 0.9708 - f1_m: 0.9733 - val_loss: 0.0042 - val_accuracy: 0.9537 - val_precision_m: 0.9581 - val_recall_m: 0.9511 - val_f1_m: 0.9545\n",
      "Epoch 94/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0026 - accuracy: 0.9726 - precision_m: 0.9748 - recall_m: 0.9700 - f1_m: 0.9724 - val_loss: 0.0039 - val_accuracy: 0.9574 - val_precision_m: 0.9593 - val_recall_m: 0.9544 - val_f1_m: 0.9568\n",
      "Epoch 95/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0024 - accuracy: 0.9749 - precision_m: 0.9778 - recall_m: 0.9726 - f1_m: 0.9751 - val_loss: 0.0041 - val_accuracy: 0.9551 - val_precision_m: 0.9576 - val_recall_m: 0.9528 - val_f1_m: 0.9552\n",
      "Epoch 96/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0025 - accuracy: 0.9734 - precision_m: 0.9751 - recall_m: 0.9707 - f1_m: 0.9729 - val_loss: 0.0038 - val_accuracy: 0.9586 - val_precision_m: 0.9610 - val_recall_m: 0.9551 - val_f1_m: 0.9580\n",
      "Epoch 97/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0024 - accuracy: 0.9745 - precision_m: 0.9765 - recall_m: 0.9720 - f1_m: 0.9742 - val_loss: 0.0042 - val_accuracy: 0.9535 - val_precision_m: 0.9575 - val_recall_m: 0.9513 - val_f1_m: 0.9544\n",
      "Epoch 98/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0024 - accuracy: 0.9738 - precision_m: 0.9761 - recall_m: 0.9719 - f1_m: 0.9740 - val_loss: 0.0039 - val_accuracy: 0.9559 - val_precision_m: 0.9583 - val_recall_m: 0.9532 - val_f1_m: 0.9557\n",
      "Epoch 99/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0024 - accuracy: 0.9747 - precision_m: 0.9768 - recall_m: 0.9718 - f1_m: 0.9742 - val_loss: 0.0039 - val_accuracy: 0.9566 - val_precision_m: 0.9596 - val_recall_m: 0.9546 - val_f1_m: 0.9570\n",
      "Epoch 100/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0024 - accuracy: 0.9738 - precision_m: 0.9757 - recall_m: 0.9715 - f1_m: 0.9736 - val_loss: 0.0038 - val_accuracy: 0.9580 - val_precision_m: 0.9616 - val_recall_m: 0.9551 - val_f1_m: 0.9583\n",
      "Epoch 101/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0024 - accuracy: 0.9749 - precision_m: 0.9778 - recall_m: 0.9724 - f1_m: 0.9751 - val_loss: 0.0038 - val_accuracy: 0.9566 - val_precision_m: 0.9595 - val_recall_m: 0.9549 - val_f1_m: 0.9572\n",
      "Epoch 102/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9755 - precision_m: 0.9778 - recall_m: 0.9730 - f1_m: 0.9754 - val_loss: 0.0041 - val_accuracy: 0.9541 - val_precision_m: 0.9585 - val_recall_m: 0.9517 - val_f1_m: 0.9550\n",
      "Epoch 103/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9757 - precision_m: 0.9780 - recall_m: 0.9739 - f1_m: 0.9759 - val_loss: 0.0041 - val_accuracy: 0.9537 - val_precision_m: 0.9573 - val_recall_m: 0.9515 - val_f1_m: 0.9544\n",
      "Epoch 104/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0023 - accuracy: 0.9758 - precision_m: 0.9780 - recall_m: 0.9733 - f1_m: 0.9756 - val_loss: 0.0038 - val_accuracy: 0.9566 - val_precision_m: 0.9601 - val_recall_m: 0.9555 - val_f1_m: 0.9578\n",
      "Epoch 105/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0022 - accuracy: 0.9778 - precision_m: 0.9793 - recall_m: 0.9753 - f1_m: 0.9772 - val_loss: 0.0041 - val_accuracy: 0.9557 - val_precision_m: 0.9584 - val_recall_m: 0.9542 - val_f1_m: 0.9562\n",
      "Epoch 106/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9759 - precision_m: 0.9780 - recall_m: 0.9739 - f1_m: 0.9759 - val_loss: 0.0040 - val_accuracy: 0.9553 - val_precision_m: 0.9596 - val_recall_m: 0.9538 - val_f1_m: 0.9567\n",
      "Epoch 107/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9759 - precision_m: 0.9775 - recall_m: 0.9747 - f1_m: 0.9761 - val_loss: 0.0037 - val_accuracy: 0.9601 - val_precision_m: 0.9628 - val_recall_m: 0.9574 - val_f1_m: 0.9601\n",
      "Epoch 108/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9758 - precision_m: 0.9776 - recall_m: 0.9738 - f1_m: 0.9757 - val_loss: 0.0041 - val_accuracy: 0.9547 - val_precision_m: 0.9578 - val_recall_m: 0.9525 - val_f1_m: 0.9551\n",
      "Epoch 109/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9746 - precision_m: 0.9770 - recall_m: 0.9725 - f1_m: 0.9747 - val_loss: 0.0037 - val_accuracy: 0.9574 - val_precision_m: 0.9595 - val_recall_m: 0.9555 - val_f1_m: 0.9575\n",
      "Epoch 110/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9770 - precision_m: 0.9790 - recall_m: 0.9749 - f1_m: 0.9769 - val_loss: 0.0034 - val_accuracy: 0.9626 - val_precision_m: 0.9655 - val_recall_m: 0.9595 - val_f1_m: 0.9625\n",
      "Epoch 111/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9782 - precision_m: 0.9808 - recall_m: 0.9765 - f1_m: 0.9786 - val_loss: 0.0038 - val_accuracy: 0.9566 - val_precision_m: 0.9601 - val_recall_m: 0.9544 - val_f1_m: 0.9572\n",
      "Epoch 112/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9775 - precision_m: 0.9792 - recall_m: 0.9764 - f1_m: 0.9778 - val_loss: 0.0041 - val_accuracy: 0.9543 - val_precision_m: 0.9574 - val_recall_m: 0.9526 - val_f1_m: 0.9550\n",
      "Epoch 113/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9778 - precision_m: 0.9796 - recall_m: 0.9760 - f1_m: 0.9778 - val_loss: 0.0038 - val_accuracy: 0.9566 - val_precision_m: 0.9597 - val_recall_m: 0.9553 - val_f1_m: 0.9575\n",
      "Epoch 114/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0023 - accuracy: 0.9765 - precision_m: 0.9784 - recall_m: 0.9748 - f1_m: 0.9766 - val_loss: 0.0037 - val_accuracy: 0.9588 - val_precision_m: 0.9620 - val_recall_m: 0.9567 - val_f1_m: 0.9593\n",
      "Epoch 115/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0022 - accuracy: 0.9760 - precision_m: 0.9781 - recall_m: 0.9734 - f1_m: 0.9757 - val_loss: 0.0037 - val_accuracy: 0.9588 - val_precision_m: 0.9609 - val_recall_m: 0.9567 - val_f1_m: 0.9587\n",
      "Epoch 116/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0020 - accuracy: 0.9778 - precision_m: 0.9795 - recall_m: 0.9765 - f1_m: 0.9780 - val_loss: 0.0035 - val_accuracy: 0.9624 - val_precision_m: 0.9636 - val_recall_m: 0.9607 - val_f1_m: 0.9621\n",
      "Epoch 117/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0020 - accuracy: 0.9796 - precision_m: 0.9810 - recall_m: 0.9778 - f1_m: 0.9794 - val_loss: 0.0035 - val_accuracy: 0.9620 - val_precision_m: 0.9650 - val_recall_m: 0.9611 - val_f1_m: 0.9630\n",
      "Epoch 118/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9777 - precision_m: 0.9795 - recall_m: 0.9757 - f1_m: 0.9776 - val_loss: 0.0038 - val_accuracy: 0.9582 - val_precision_m: 0.9610 - val_recall_m: 0.9569 - val_f1_m: 0.9589\n",
      "Epoch 119/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0020 - accuracy: 0.9794 - precision_m: 0.9810 - recall_m: 0.9773 - f1_m: 0.9791 - val_loss: 0.0036 - val_accuracy: 0.9586 - val_precision_m: 0.9618 - val_recall_m: 0.9572 - val_f1_m: 0.9595\n",
      "Epoch 120/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0020 - accuracy: 0.9785 - precision_m: 0.9803 - recall_m: 0.9770 - f1_m: 0.9786 - val_loss: 0.0036 - val_accuracy: 0.9593 - val_precision_m: 0.9621 - val_recall_m: 0.9572 - val_f1_m: 0.9596\n",
      "Epoch 121/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9774 - precision_m: 0.9794 - recall_m: 0.9762 - f1_m: 0.9778 - val_loss: 0.0036 - val_accuracy: 0.9609 - val_precision_m: 0.9631 - val_recall_m: 0.9597 - val_f1_m: 0.9614\n",
      "Epoch 122/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0019 - accuracy: 0.9800 - precision_m: 0.9813 - recall_m: 0.9781 - f1_m: 0.9797 - val_loss: 0.0035 - val_accuracy: 0.9614 - val_precision_m: 0.9633 - val_recall_m: 0.9601 - val_f1_m: 0.9617\n",
      "Epoch 123/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0021 - accuracy: 0.9765 - precision_m: 0.9787 - recall_m: 0.9752 - f1_m: 0.9769 - val_loss: 0.0040 - val_accuracy: 0.9545 - val_precision_m: 0.9583 - val_recall_m: 0.9519 - val_f1_m: 0.9550\n",
      "Epoch 124/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0021 - accuracy: 0.9780 - precision_m: 0.9800 - recall_m: 0.9770 - f1_m: 0.9785 - val_loss: 0.0032 - val_accuracy: 0.9628 - val_precision_m: 0.9654 - val_recall_m: 0.9617 - val_f1_m: 0.9635\n",
      "Epoch 125/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0019 - accuracy: 0.9805 - precision_m: 0.9823 - recall_m: 0.9791 - f1_m: 0.9807 - val_loss: 0.0031 - val_accuracy: 0.9655 - val_precision_m: 0.9672 - val_recall_m: 0.9640 - val_f1_m: 0.9656\n",
      "Epoch 126/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0018 - accuracy: 0.9816 - precision_m: 0.9824 - recall_m: 0.9809 - f1_m: 0.9816 - val_loss: 0.0038 - val_accuracy: 0.9562 - val_precision_m: 0.9593 - val_recall_m: 0.9551 - val_f1_m: 0.9572\n",
      "Epoch 127/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0020 - accuracy: 0.9795 - precision_m: 0.9809 - recall_m: 0.9779 - f1_m: 0.9794 - val_loss: 0.0038 - val_accuracy: 0.9580 - val_precision_m: 0.9605 - val_recall_m: 0.9571 - val_f1_m: 0.9588\n",
      "Epoch 128/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0018 - accuracy: 0.9813 - precision_m: 0.9828 - recall_m: 0.9799 - f1_m: 0.9813 - val_loss: 0.0038 - val_accuracy: 0.9568 - val_precision_m: 0.9593 - val_recall_m: 0.9551 - val_f1_m: 0.9572\n",
      "Epoch 129/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0020 - accuracy: 0.9792 - precision_m: 0.9806 - recall_m: 0.9775 - f1_m: 0.9790 - val_loss: 0.0037 - val_accuracy: 0.9605 - val_precision_m: 0.9628 - val_recall_m: 0.9586 - val_f1_m: 0.9607\n",
      "Epoch 130/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0019 - accuracy: 0.9803 - precision_m: 0.9819 - recall_m: 0.9791 - f1_m: 0.9805 - val_loss: 0.0038 - val_accuracy: 0.9586 - val_precision_m: 0.9612 - val_recall_m: 0.9565 - val_f1_m: 0.9588\n",
      "Epoch 131/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9825 - precision_m: 0.9831 - recall_m: 0.9809 - f1_m: 0.9820 - val_loss: 0.0031 - val_accuracy: 0.9663 - val_precision_m: 0.9693 - val_recall_m: 0.9636 - val_f1_m: 0.9664\n",
      "Epoch 132/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0018 - accuracy: 0.9819 - precision_m: 0.9833 - recall_m: 0.9807 - f1_m: 0.9820 - val_loss: 0.0030 - val_accuracy: 0.9667 - val_precision_m: 0.9683 - val_recall_m: 0.9653 - val_f1_m: 0.9668\n",
      "Epoch 133/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0017 - accuracy: 0.9818 - precision_m: 0.9834 - recall_m: 0.9806 - f1_m: 0.9820 - val_loss: 0.0038 - val_accuracy: 0.9576 - val_precision_m: 0.9613 - val_recall_m: 0.9544 - val_f1_m: 0.9578\n",
      "Epoch 134/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0017 - accuracy: 0.9821 - precision_m: 0.9833 - recall_m: 0.9809 - f1_m: 0.9821 - val_loss: 0.0038 - val_accuracy: 0.9570 - val_precision_m: 0.9608 - val_recall_m: 0.9551 - val_f1_m: 0.9579\n",
      "Epoch 135/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9823 - precision_m: 0.9842 - recall_m: 0.9807 - f1_m: 0.9824 - val_loss: 0.0033 - val_accuracy: 0.9632 - val_precision_m: 0.9653 - val_recall_m: 0.9607 - val_f1_m: 0.9630\n",
      "Epoch 136/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0018 - accuracy: 0.9810 - precision_m: 0.9830 - recall_m: 0.9800 - f1_m: 0.9815 - val_loss: 0.0035 - val_accuracy: 0.9630 - val_precision_m: 0.9644 - val_recall_m: 0.9613 - val_f1_m: 0.9628\n",
      "Epoch 137/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0017 - accuracy: 0.9824 - precision_m: 0.9837 - recall_m: 0.9811 - f1_m: 0.9824 - val_loss: 0.0029 - val_accuracy: 0.9674 - val_precision_m: 0.9701 - val_recall_m: 0.9657 - val_f1_m: 0.9679\n",
      "Epoch 138/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9816 - precision_m: 0.9832 - recall_m: 0.9801 - f1_m: 0.9816 - val_loss: 0.0034 - val_accuracy: 0.9632 - val_precision_m: 0.9652 - val_recall_m: 0.9603 - val_f1_m: 0.9627\n",
      "Epoch 139/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9823 - precision_m: 0.9837 - recall_m: 0.9811 - f1_m: 0.9824 - val_loss: 0.0036 - val_accuracy: 0.9605 - val_precision_m: 0.9621 - val_recall_m: 0.9595 - val_f1_m: 0.9608\n",
      "Epoch 140/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0017 - accuracy: 0.9827 - precision_m: 0.9840 - recall_m: 0.9816 - f1_m: 0.9828 - val_loss: 0.0030 - val_accuracy: 0.9659 - val_precision_m: 0.9686 - val_recall_m: 0.9653 - val_f1_m: 0.9669\n",
      "Epoch 141/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0016 - accuracy: 0.9839 - precision_m: 0.9847 - recall_m: 0.9828 - f1_m: 0.9837 - val_loss: 0.0030 - val_accuracy: 0.9649 - val_precision_m: 0.9673 - val_recall_m: 0.9641 - val_f1_m: 0.9657\n",
      "Epoch 142/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9824 - precision_m: 0.9837 - recall_m: 0.9813 - f1_m: 0.9825 - val_loss: 0.0032 - val_accuracy: 0.9643 - val_precision_m: 0.9675 - val_recall_m: 0.9626 - val_f1_m: 0.9650\n",
      "Epoch 143/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9821 - precision_m: 0.9835 - recall_m: 0.9809 - f1_m: 0.9821 - val_loss: 0.0033 - val_accuracy: 0.9626 - val_precision_m: 0.9648 - val_recall_m: 0.9611 - val_f1_m: 0.9629\n",
      "Epoch 144/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0015 - accuracy: 0.9830 - precision_m: 0.9841 - recall_m: 0.9822 - f1_m: 0.9831 - val_loss: 0.0032 - val_accuracy: 0.9655 - val_precision_m: 0.9673 - val_recall_m: 0.9636 - val_f1_m: 0.9654\n",
      "Epoch 145/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0015 - accuracy: 0.9839 - precision_m: 0.9853 - recall_m: 0.9830 - f1_m: 0.9841 - val_loss: 0.0031 - val_accuracy: 0.9657 - val_precision_m: 0.9690 - val_recall_m: 0.9645 - val_f1_m: 0.9667\n",
      "Epoch 146/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0014 - accuracy: 0.9855 - precision_m: 0.9863 - recall_m: 0.9847 - f1_m: 0.9854 - val_loss: 0.0035 - val_accuracy: 0.9601 - val_precision_m: 0.9628 - val_recall_m: 0.9586 - val_f1_m: 0.9607\n",
      "Epoch 147/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0016 - accuracy: 0.9841 - precision_m: 0.9850 - recall_m: 0.9831 - f1_m: 0.9840 - val_loss: 0.0032 - val_accuracy: 0.9632 - val_precision_m: 0.9648 - val_recall_m: 0.9613 - val_f1_m: 0.9630\n",
      "Epoch 148/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0014 - accuracy: 0.9856 - precision_m: 0.9867 - recall_m: 0.9846 - f1_m: 0.9856 - val_loss: 0.0031 - val_accuracy: 0.9651 - val_precision_m: 0.9675 - val_recall_m: 0.9628 - val_f1_m: 0.9651\n",
      "Epoch 149/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0016 - accuracy: 0.9834 - precision_m: 0.9846 - recall_m: 0.9824 - f1_m: 0.9834 - val_loss: 0.0030 - val_accuracy: 0.9670 - val_precision_m: 0.9696 - val_recall_m: 0.9663 - val_f1_m: 0.9679\n",
      "Epoch 150/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0015 - accuracy: 0.9844 - precision_m: 0.9855 - recall_m: 0.9834 - f1_m: 0.9844 - val_loss: 0.0035 - val_accuracy: 0.9611 - val_precision_m: 0.9632 - val_recall_m: 0.9599 - val_f1_m: 0.9616\n",
      "Epoch 151/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0015 - accuracy: 0.9845 - precision_m: 0.9851 - recall_m: 0.9838 - f1_m: 0.9844 - val_loss: 0.0031 - val_accuracy: 0.9659 - val_precision_m: 0.9671 - val_recall_m: 0.9645 - val_f1_m: 0.9658\n",
      "Epoch 152/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0017 - accuracy: 0.9810 - precision_m: 0.9825 - recall_m: 0.9797 - f1_m: 0.9811 - val_loss: 0.0031 - val_accuracy: 0.9634 - val_precision_m: 0.9671 - val_recall_m: 0.9615 - val_f1_m: 0.9643\n",
      "Epoch 153/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0016 - accuracy: 0.9829 - precision_m: 0.9842 - recall_m: 0.9813 - f1_m: 0.9827 - val_loss: 0.0032 - val_accuracy: 0.9649 - val_precision_m: 0.9673 - val_recall_m: 0.9641 - val_f1_m: 0.9657\n",
      "Epoch 154/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0013 - accuracy: 0.9863 - precision_m: 0.9877 - recall_m: 0.9855 - f1_m: 0.9866 - val_loss: 0.0033 - val_accuracy: 0.9655 - val_precision_m: 0.9671 - val_recall_m: 0.9643 - val_f1_m: 0.9657\n",
      "Epoch 155/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0012 - accuracy: 0.9873 - precision_m: 0.9881 - recall_m: 0.9867 - f1_m: 0.9874 - val_loss: 0.0031 - val_accuracy: 0.9653 - val_precision_m: 0.9673 - val_recall_m: 0.9641 - val_f1_m: 0.9657\n",
      "Epoch 156/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0013 - accuracy: 0.9864 - precision_m: 0.9879 - recall_m: 0.9856 - f1_m: 0.9867 - val_loss: 0.0027 - val_accuracy: 0.9690 - val_precision_m: 0.9704 - val_recall_m: 0.9682 - val_f1_m: 0.9693\n",
      "Epoch 157/200\n",
      "487/487 [==============================] - 4s 9ms/step - loss: 0.0012 - accuracy: 0.9879 - precision_m: 0.9886 - recall_m: 0.9870 - f1_m: 0.9878 - val_loss: 0.0030 - val_accuracy: 0.9667 - val_precision_m: 0.9687 - val_recall_m: 0.9657 - val_f1_m: 0.9671\n",
      "Epoch 158/200\n",
      "487/487 [==============================] - 4s 9ms/step - loss: 0.0012 - accuracy: 0.9870 - precision_m: 0.9882 - recall_m: 0.9861 - f1_m: 0.9872 - val_loss: 0.0029 - val_accuracy: 0.9668 - val_precision_m: 0.9691 - val_recall_m: 0.9664 - val_f1_m: 0.9677\n",
      "Epoch 159/200\n",
      "487/487 [==============================] - 4s 9ms/step - loss: 0.0012 - accuracy: 0.9879 - precision_m: 0.9888 - recall_m: 0.9870 - f1_m: 0.9879 - val_loss: 0.0028 - val_accuracy: 0.9684 - val_precision_m: 0.9700 - val_recall_m: 0.9668 - val_f1_m: 0.9684\n",
      "Epoch 160/200\n",
      "487/487 [==============================] - 4s 9ms/step - loss: 0.0012 - accuracy: 0.9875 - precision_m: 0.9884 - recall_m: 0.9867 - f1_m: 0.9876 - val_loss: 0.0027 - val_accuracy: 0.9695 - val_precision_m: 0.9717 - val_recall_m: 0.9678 - val_f1_m: 0.9697\n",
      "Epoch 161/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0011 - accuracy: 0.9888 - precision_m: 0.9899 - recall_m: 0.9881 - f1_m: 0.9890 - val_loss: 0.0029 - val_accuracy: 0.9676 - val_precision_m: 0.9702 - val_recall_m: 0.9672 - val_f1_m: 0.9687\n",
      "Epoch 162/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9870 - precision_m: 0.9878 - recall_m: 0.9860 - f1_m: 0.9869 - val_loss: 0.0030 - val_accuracy: 0.9668 - val_precision_m: 0.9692 - val_recall_m: 0.9651 - val_f1_m: 0.9671\n",
      "Epoch 163/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9880 - precision_m: 0.9890 - recall_m: 0.9870 - f1_m: 0.9880 - val_loss: 0.0030 - val_accuracy: 0.9659 - val_precision_m: 0.9681 - val_recall_m: 0.9638 - val_f1_m: 0.9659\n",
      "Epoch 164/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9880 - precision_m: 0.9891 - recall_m: 0.9874 - f1_m: 0.9882 - val_loss: 0.0029 - val_accuracy: 0.9676 - val_precision_m: 0.9690 - val_recall_m: 0.9668 - val_f1_m: 0.9679\n",
      "Epoch 165/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9880 - precision_m: 0.9886 - recall_m: 0.9873 - f1_m: 0.9879 - val_loss: 0.0031 - val_accuracy: 0.9636 - val_precision_m: 0.9663 - val_recall_m: 0.9626 - val_f1_m: 0.9644\n",
      "Epoch 166/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9870 - precision_m: 0.9878 - recall_m: 0.9863 - f1_m: 0.9870 - val_loss: 0.0030 - val_accuracy: 0.9641 - val_precision_m: 0.9663 - val_recall_m: 0.9624 - val_f1_m: 0.9643\n",
      "Epoch 167/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9872 - precision_m: 0.9886 - recall_m: 0.9867 - f1_m: 0.9876 - val_loss: 0.0031 - val_accuracy: 0.9655 - val_precision_m: 0.9675 - val_recall_m: 0.9651 - val_f1_m: 0.9663\n",
      "Epoch 168/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0011 - accuracy: 0.9875 - precision_m: 0.9886 - recall_m: 0.9868 - f1_m: 0.9877 - val_loss: 0.0029 - val_accuracy: 0.9680 - val_precision_m: 0.9694 - val_recall_m: 0.9670 - val_f1_m: 0.9682\n",
      "Epoch 169/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9880 - precision_m: 0.9887 - recall_m: 0.9873 - f1_m: 0.9880 - val_loss: 0.0030 - val_accuracy: 0.9670 - val_precision_m: 0.9688 - val_recall_m: 0.9659 - val_f1_m: 0.9673\n",
      "Epoch 170/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9865 - precision_m: 0.9872 - recall_m: 0.9859 - f1_m: 0.9866 - val_loss: 0.0028 - val_accuracy: 0.9684 - val_precision_m: 0.9707 - val_recall_m: 0.9663 - val_f1_m: 0.9684\n",
      "Epoch 171/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9883 - precision_m: 0.9887 - recall_m: 0.9878 - f1_m: 0.9882 - val_loss: 0.0031 - val_accuracy: 0.9653 - val_precision_m: 0.9677 - val_recall_m: 0.9640 - val_f1_m: 0.9658\n",
      "Epoch 172/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9871 - precision_m: 0.9880 - recall_m: 0.9863 - f1_m: 0.9871 - val_loss: 0.0029 - val_accuracy: 0.9682 - val_precision_m: 0.9710 - val_recall_m: 0.9674 - val_f1_m: 0.9692\n",
      "Epoch 173/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9889 - precision_m: 0.9894 - recall_m: 0.9886 - f1_m: 0.9890 - val_loss: 0.0028 - val_accuracy: 0.9695 - val_precision_m: 0.9711 - val_recall_m: 0.9686 - val_f1_m: 0.9698\n",
      "Epoch 174/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9880 - precision_m: 0.9891 - recall_m: 0.9874 - f1_m: 0.9883 - val_loss: 0.0029 - val_accuracy: 0.9676 - val_precision_m: 0.9698 - val_recall_m: 0.9661 - val_f1_m: 0.9679\n",
      "Epoch 175/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0010 - accuracy: 0.9891 - precision_m: 0.9895 - recall_m: 0.9881 - f1_m: 0.9888 - val_loss: 0.0030 - val_accuracy: 0.9653 - val_precision_m: 0.9671 - val_recall_m: 0.9643 - val_f1_m: 0.9657\n",
      "Epoch 176/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0011 - accuracy: 0.9881 - precision_m: 0.9891 - recall_m: 0.9877 - f1_m: 0.9884 - val_loss: 0.0028 - val_accuracy: 0.9663 - val_precision_m: 0.9700 - val_recall_m: 0.9653 - val_f1_m: 0.9676\n",
      "Epoch 177/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0012 - accuracy: 0.9877 - precision_m: 0.9881 - recall_m: 0.9868 - f1_m: 0.9874 - val_loss: 0.0031 - val_accuracy: 0.9649 - val_precision_m: 0.9669 - val_recall_m: 0.9632 - val_f1_m: 0.9650\n",
      "Epoch 178/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0013 - accuracy: 0.9862 - precision_m: 0.9873 - recall_m: 0.9859 - f1_m: 0.9866 - val_loss: 0.0029 - val_accuracy: 0.9676 - val_precision_m: 0.9705 - val_recall_m: 0.9663 - val_f1_m: 0.9684\n",
      "Epoch 179/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0012 - accuracy: 0.9875 - precision_m: 0.9877 - recall_m: 0.9867 - f1_m: 0.9872 - val_loss: 0.0034 - val_accuracy: 0.9630 - val_precision_m: 0.9648 - val_recall_m: 0.9615 - val_f1_m: 0.9631\n",
      "Epoch 180/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0013 - accuracy: 0.9870 - precision_m: 0.9879 - recall_m: 0.9863 - f1_m: 0.9871 - val_loss: 0.0028 - val_accuracy: 0.9690 - val_precision_m: 0.9714 - val_recall_m: 0.9682 - val_f1_m: 0.9697\n",
      "Epoch 181/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0010 - accuracy: 0.9897 - precision_m: 0.9902 - recall_m: 0.9890 - f1_m: 0.9896 - val_loss: 0.0029 - val_accuracy: 0.9688 - val_precision_m: 0.9699 - val_recall_m: 0.9678 - val_f1_m: 0.9688\n",
      "Epoch 182/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0010 - accuracy: 0.9889 - precision_m: 0.9896 - recall_m: 0.9885 - f1_m: 0.9890 - val_loss: 0.0035 - val_accuracy: 0.9630 - val_precision_m: 0.9646 - val_recall_m: 0.9615 - val_f1_m: 0.9630\n",
      "Epoch 183/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0011 - accuracy: 0.9885 - precision_m: 0.9891 - recall_m: 0.9880 - f1_m: 0.9885 - val_loss: 0.0027 - val_accuracy: 0.9699 - val_precision_m: 0.9714 - val_recall_m: 0.9695 - val_f1_m: 0.9704\n",
      "Epoch 184/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9876 - precision_m: 0.9882 - recall_m: 0.9873 - f1_m: 0.9877 - val_loss: 0.0029 - val_accuracy: 0.9688 - val_precision_m: 0.9704 - val_recall_m: 0.9680 - val_f1_m: 0.9692\n",
      "Epoch 185/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9880 - precision_m: 0.9889 - recall_m: 0.9874 - f1_m: 0.9881 - val_loss: 0.0029 - val_accuracy: 0.9690 - val_precision_m: 0.9696 - val_recall_m: 0.9680 - val_f1_m: 0.9688\n",
      "Epoch 186/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9863 - precision_m: 0.9869 - recall_m: 0.9858 - f1_m: 0.9863 - val_loss: 0.0027 - val_accuracy: 0.9688 - val_precision_m: 0.9710 - val_recall_m: 0.9680 - val_f1_m: 0.9695\n",
      "Epoch 187/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9859 - precision_m: 0.9863 - recall_m: 0.9851 - f1_m: 0.9857 - val_loss: 0.0034 - val_accuracy: 0.9641 - val_precision_m: 0.9654 - val_recall_m: 0.9638 - val_f1_m: 0.9646\n",
      "Epoch 188/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9866 - precision_m: 0.9876 - recall_m: 0.9861 - f1_m: 0.9869 - val_loss: 0.0032 - val_accuracy: 0.9641 - val_precision_m: 0.9665 - val_recall_m: 0.9622 - val_f1_m: 0.9643\n",
      "Epoch 189/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9873 - precision_m: 0.9879 - recall_m: 0.9870 - f1_m: 0.9874 - val_loss: 0.0032 - val_accuracy: 0.9647 - val_precision_m: 0.9665 - val_recall_m: 0.9636 - val_f1_m: 0.9650\n",
      "Epoch 190/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9888 - precision_m: 0.9894 - recall_m: 0.9884 - f1_m: 0.9889 - val_loss: 0.0033 - val_accuracy: 0.9628 - val_precision_m: 0.9652 - val_recall_m: 0.9620 - val_f1_m: 0.9636\n",
      "Epoch 191/200\n",
      "487/487 [==============================] - 5s 9ms/step - loss: 0.0012 - accuracy: 0.9879 - precision_m: 0.9885 - recall_m: 0.9874 - f1_m: 0.9880 - val_loss: 0.0035 - val_accuracy: 0.9624 - val_precision_m: 0.9644 - val_recall_m: 0.9607 - val_f1_m: 0.9625\n",
      "Epoch 192/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9873 - precision_m: 0.9879 - recall_m: 0.9868 - f1_m: 0.9873 - val_loss: 0.0033 - val_accuracy: 0.9640 - val_precision_m: 0.9654 - val_recall_m: 0.9632 - val_f1_m: 0.9643\n",
      "Epoch 193/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9867 - precision_m: 0.9872 - recall_m: 0.9861 - f1_m: 0.9866 - val_loss: 0.0031 - val_accuracy: 0.9665 - val_precision_m: 0.9684 - val_recall_m: 0.9655 - val_f1_m: 0.9669\n",
      "Epoch 194/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9867 - precision_m: 0.9875 - recall_m: 0.9862 - f1_m: 0.9868 - val_loss: 0.0032 - val_accuracy: 0.9663 - val_precision_m: 0.9687 - val_recall_m: 0.9657 - val_f1_m: 0.9671\n",
      "Epoch 195/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9863 - precision_m: 0.9867 - recall_m: 0.9859 - f1_m: 0.9863 - val_loss: 0.0033 - val_accuracy: 0.9651 - val_precision_m: 0.9673 - val_recall_m: 0.9636 - val_f1_m: 0.9654\n",
      "Epoch 196/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9882 - precision_m: 0.9891 - recall_m: 0.9881 - f1_m: 0.9886 - val_loss: 0.0033 - val_accuracy: 0.9638 - val_precision_m: 0.9650 - val_recall_m: 0.9628 - val_f1_m: 0.9639\n",
      "Epoch 197/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0011 - accuracy: 0.9889 - precision_m: 0.9895 - recall_m: 0.9883 - f1_m: 0.9889 - val_loss: 0.0031 - val_accuracy: 0.9663 - val_precision_m: 0.9681 - val_recall_m: 0.9655 - val_f1_m: 0.9668\n",
      "Epoch 198/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0012 - accuracy: 0.9884 - precision_m: 0.9888 - recall_m: 0.9881 - f1_m: 0.9884 - val_loss: 0.0030 - val_accuracy: 0.9678 - val_precision_m: 0.9698 - val_recall_m: 0.9663 - val_f1_m: 0.9680\n",
      "Epoch 199/200\n",
      "487/487 [==============================] - 5s 10ms/step - loss: 0.0013 - accuracy: 0.9873 - precision_m: 0.9878 - recall_m: 0.9866 - f1_m: 0.9872 - val_loss: 0.0030 - val_accuracy: 0.9678 - val_precision_m: 0.9693 - val_recall_m: 0.9668 - val_f1_m: 0.9680\n",
      "Epoch 200/200\n",
      "487/487 [==============================] - 5s 11ms/step - loss: 0.0011 - accuracy: 0.9888 - precision_m: 0.9893 - recall_m: 0.9882 - f1_m: 0.9888 - val_loss: 0.0029 - val_accuracy: 0.9686 - val_precision_m: 0.9702 - val_recall_m: 0.9682 - val_f1_m: 0.9692\n"
     ]
    }
   ],
   "source": [
    "history= model.fit(X_train, y_train, epochs=200,batch_size=32, validation_data=(X_test, y_test))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [],
   "source": [
    "del model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "id": "gsUnLWTHATbK"
   },
   "outputs": [],
   "source": [
    "train_loss = history.history['loss']\n",
    "valid_loss = history.history['val_loss']\n",
    "\n",
    "train_acc = history.history['accuracy']\n",
    "valid_acc = history.history['val_accuracy']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy import savetxt\n",
    "savetxt('train_loss_LLD.csv', train_loss, delimiter=',')\n",
    "savetxt('valid_loss_LLD.csv', valid_loss, delimiter=',')\n",
    "\n",
    "savetxt('train_acc_LLD.csv', train_acc, delimiter=',')\n",
    "savetxt('valid_acc_LLD.csv', valid_acc, delimiter=',')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "fig, axes = plt.subplots(1, 2, figsize=(13,4))\n",
    "axes = axes.flatten()\n",
    "\n",
    "axes[0].plot(train_acc, label='training')\n",
    "axes[0].plot(valid_acc, label='validation')\n",
    "axes[0].set_title('Accuracy Curve')\n",
    "axes[0].set_xlabel('epochs')\n",
    "axes[0].set_ylabel('accuracy')\n",
    "axes[0].legend()\n",
    "\n",
    "\n",
    "axes[1].plot(train_loss, label='training')\n",
    "axes[1].plot(valid_loss, label='validation')\n",
    "axes[1].set_title('Loss Curve')\n",
    "axes[1].set_xlabel('epochs')\n",
    "axes[1].set_ylabel('loss')\n",
    "axes[1].legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_f1 = history.history['f1_m']\n",
    "valid_f1 = history.history['val_f1_m']\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy import savetxt\n",
    "savetxt('train_f1_m.csv', train_f1, delimiter=',')\n",
    "savetxt('valid_f1_m.csv', valid_f1, delimiter=',')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(1, 2, figsize=(20,7))\n",
    "axes = axes.flatten()\n",
    "\n",
    "\n",
    "\n",
    "axes[1].plot(train_f1, label='training')\n",
    "axes[1].plot(valid_f1, label='validation')\n",
    "axes[1].set_title('f1 Curve')\n",
    "axes[1].set_xlabel('epochs')\n",
    "axes[1].set_ylabel('dice_coef')\n",
    "axes[1].legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "163/163 [==============================] - 1s 3ms/step\n"
     ]
    }
   ],
   "source": [
    "y_predict=model.predict(X_test)\n",
    "y_pred=np.argmax(y_predict,axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test=np.argmax(y_test,axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[459   8   3   1   0   1   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  6 455   7   0   0   1   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  4   3 438   2   1   1   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0 462   0   0   0   0   0   0   0   1   0   2   1   1   0   0]\n",
      " [  0   0   1   1 540   1   0   1   0   0   1   0   0   0   0   0   0   0]\n",
      " [  5   1   2   0   0 445   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   3   4   1   1 426   4   0   0   2   0   0   0   0   0   0   0]\n",
      " [  0   0   0   2   7   0   5 317   0   0   0   0   0   0   0   1   0   1]\n",
      " [  0   0   0   0   0   0   0   0 166   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   1   0 118   1   0   0   0   0   0   0   0]\n",
      " [  0   0   1   0  10   0   3   0   0   0 236   0   0   0   0   1   0   0]\n",
      " [  0   0   0   2   0   0   0   0   0   0   0 210   7   1   0   1   0   0]\n",
      " [  1   0   0   3   0   0   0   0   0   0   0   1  74   0   0   0   0   0]\n",
      " [  0   0   0   1   0   0   0   0   0   0   0   2   0  61   0   1   0   0]\n",
      " [  0   0   0   0   0   0   0   7   0  13   0   0   0   0 129   0   0   0]\n",
      " [  0   0   0   1   0   0   0   1   0   0   0   0   0   4   0 190   3   0]\n",
      " [  0   0   0   1   0   0   0   0   0   0   0   2   1   0   0   5 186   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   1 113]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "confusion_matrix = metrics.confusion_matrix(y_test, y_pred)\n",
    "cm_display = metrics.ConfusionMatrixDisplay(confusion_matrix = confusion_matrix, display_labels = [False, True])\n",
    "print(confusion_matrix)\n",
    "savetxt('confusion_matrix.csv', confusion_matrix, delimiter=',')\n",
    "\n"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "provenance": []
  },
  "gpuClass": "standard",
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}