Train_CNN_Digital-Readout-Small-v2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CNN Training\n",
    "\n",
    "Target of this code is to train a CNN network to classify images of a digital readout to the digits 0 to 9. Additionally a category \"NaN\" is introduced, to mark images that are not amibiguous.\n",
    "\n",
    "### Preparing the training\n",
    "* First all libraries are loaded\n",
    "    * It is assumed, that they are installed during the Python setup\n",
    "* matplotlib is set to print the output inline in the jupyter notebook"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "########### Basic Parameters for Running: ################################\n",
    "    \n",
    "TFlite_Version  = \"1801\"   \n",
    "TFlite_MainType = \"dig-class11\"\n",
    "TFlite_Size     = \"s2\"\n",
    "Training_Percentage = 0.0              # 0.0 = Use all Images for Training\n",
    "Epoch_Anz = 500\n",
    "\n",
    "##########################################################################\n",
    "\n",
    "## 2024-03-30: Code adapted to TF 2.16 ###################################\n",
    "\n",
    "\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import glob\n",
    "import os\n",
    "import numpy as np\n",
    "from sklearn.utils import shuffle\n",
    "from tensorflow.python import keras\n",
    "from tensorflow.python.keras import Sequential\n",
    "from tensorflow.keras.layers import Dense, InputLayer, Conv2D, MaxPool2D, Flatten, BatchNormalization\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras.callbacks import History \n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from PIL import Image \n",
    "from pathlib import Path\n",
    "\n",
    "loss_ges = np.array([])\n",
    "val_loss_ges = np.array([])\n",
    "\n",
    "%matplotlib inline\n",
    "np.set_printoptions(precision=4)\n",
    "np.set_printoptions(suppress=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load training data\n",
    "* The data is expected in the \"Input_dir\"\n",
    "* Inside subdirectories are expected from -1, 0, 1, ... 9 in which the pictures are sorted according to their values (=category)\n",
    "* Picture size must be 20x32 with 3 color channels (RGB)\n",
    "* The filename can be arbitrary\n",
    "\n",
    "* The images are stored in the x_data[]\n",
    "* The expected category for each image in the corresponding y_data[]\n",
    "\n",
    "* The last step is a shuffle (from sklearn.utils) and split the data into training and validation data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1379, 32, 20, 3)\n",
      "(1379, 11)\n"
     ]
    }
   ],
   "source": [
    "Input_dir='ziffer_sortiert_resize'\n",
    "\n",
    "files = glob.glob(Input_dir + '/*.jpg')\n",
    "x_data = []\n",
    "y_data = []\n",
    "\n",
    "for aktfile in files:\n",
    "    base = os.path.basename(aktfile)\n",
    "    target = base[0:1]\n",
    "    if target == \"N\":\n",
    "        category = 10                # NaN does not work --> convert to 10\n",
    "    else:\n",
    "        category = int(target)\n",
    "    test_image = Image.open(aktfile)\n",
    "    test_image = np.array(test_image, dtype=\"float32\")\n",
    "    x_data.append(test_image)\n",
    "    y_data.append(np.array([category]))\n",
    "\n",
    "x_data = np.array(x_data)\n",
    "y_data = np.array(y_data)\n",
    "y_data = to_categorical(y_data, 11)\n",
    "print(x_data.shape)\n",
    "print(y_data.shape)\n",
    "\n",
    "x_data, y_data = shuffle(x_data, y_data)\n",
    "\n",
    "if (Training_Percentage > 0):\n",
    "    X_train, X_test, y_train, y_test = train_test_split(x_data, y_data, test_size=Training_Percentage)\n",
    "else:\n",
    "    X_train = x_data\n",
    "    y_train = y_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define the model\n",
    "\n",
    "The layout of the network ist a typcial CNN network with alternating **Conv2D** and **MaxPool2D** layers. Finished after **flattening** with additional **Dense** layer.\n",
    "\n",
    "#### Important\n",
    "* Shape of the input layer: (32, 20, 3)\n",
    "* Number of output layers: 11\n",
    "* As loss function \"categorical_crossentropy\" is choosen, as it is a categories task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:absl:At this time, the v2.11+ optimizer `tf.keras.optimizers.Adadelta` runs slowly on M1/M2 Macs, please use the legacy Keras optimizer instead, located at `tf.keras.optimizers.legacy.Adadelta`.\n",
      "WARNING:absl:There is a known slowdown when using v2.11+ Keras optimizers on M1/M2 Macs. Falling back to the legacy Keras optimizer, i.e., `tf.keras.optimizers.legacy.Adadelta`.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_2 (InputLayer)        [(None, 32, 20, 3)]       0         \n",
      "                                                                 \n",
      " batch_normalization_1 (Bat  (None, 32, 20, 3)         12        \n",
      " chNormalization)                                                \n",
      "                                                                 \n",
      " conv2d_3 (Conv2D)           (None, 32, 20, 32)        896       \n",
      "                                                                 \n",
      " max_pooling2d_3 (MaxPoolin  (None, 16, 10, 32)        0         \n",
      " g2D)                                                            \n",
      "                                                                 \n",
      " conv2d_4 (Conv2D)           (None, 16, 10, 32)        9248      \n",
      "                                                                 \n",
      " max_pooling2d_4 (MaxPoolin  (None, 8, 5, 32)          0         \n",
      " g2D)                                                            \n",
      "                                                                 \n",
      " conv2d_5 (Conv2D)           (None, 8, 5, 32)          9248      \n",
      "                                                                 \n",
      " max_pooling2d_5 (MaxPoolin  (None, 4, 2, 32)          0         \n",
      " g2D)                                                            \n",
      "                                                                 \n",
      " flatten_1 (Flatten)         (None, 256)               0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 256)               65792     \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 11)                2827      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 88023 (343.84 KB)\n",
      "Trainable params: 88017 (343.82 KB)\n",
      "Non-trainable params: 6 (24.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "inputs = tf.keras.Input(shape=(32, 20, 3))\n",
    "inputs2 = tf.keras.layers.BatchNormalization()(inputs)\n",
    "inputs3 = tf.keras.layers.Conv2D(32, (3, 3), padding='same', activation=\"relu\")(inputs2)\n",
    "inputs4 = tf.keras.layers.MaxPool2D(pool_size=(2,2))(inputs3)\n",
    "inputs5 = tf.keras.layers.Conv2D(32, (3, 3), padding='same', activation=\"relu\")(inputs4)\n",
    "inputs6 = tf.keras.layers.MaxPool2D(pool_size=(2,2))(inputs5)\n",
    "inputs7 = tf.keras.layers.Conv2D(32, (3, 3), padding='same', activation=\"relu\")(inputs6)\n",
    "inputs8 = tf.keras.layers.MaxPool2D(pool_size=(2,2))(inputs7)\n",
    "inputs9 = tf.keras.layers.Flatten()(inputs8)\n",
    "inputs10 = tf.keras.layers.Dense(256,activation=\"relu\")(inputs9)\n",
    "output = tf.keras.layers.Dense(11, activation='softmax')(inputs10)\n",
    "\n",
    "model = tf.keras.Model(inputs=inputs, outputs=output)\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy, optimizer=tf.keras.optimizers.Adadelta(learning_rate=1.0, rho=0.95), metrics = [\"accuracy\"])\n",
    "\n",
    "model.summary()\n",
    "\n",
    "#model = Sequential()\n",
    "#model.add(BatchNormalization(input_shape=(32,20,3)))\n",
    "#model.add(Conv2D(32, (3, 3), padding='same', activation=\"relu\"))\n",
    "#model.add(MaxPool2D(pool_size=(2,2)))\n",
    "#model.add(Conv2D(32, (3, 3), padding='same', activation=\"relu\"))\n",
    "#model.add(MaxPool2D(pool_size=(2,2)))\n",
    "#model.add(Conv2D(32, (3, 3), padding='same', activation=\"relu\"))\n",
    "#model.add(MaxPool2D(pool_size=(2,2)))\n",
    "#model.add(Flatten())\n",
    "#model.add(Dense(256,activation=\"relu\"))\n",
    "#model.add(Dense(11, activation = \"softmax\"))\n",
    "\n",
    "#model.compile(loss=keras.losses.categorical_crossentropy, optimizer=tf.keras.optimizers.Adadelta(learning_rate=1.0, rho=0.95), metrics = [\"accuracy\"])\n",
    "\n",
    "#model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training\n",
    "The input pictures are randomly scattered for brightness, pixel shift variations and rotation angle. This is implemented with a ImageDataGenerator.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/500\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-07-21 16:17:55.310101: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "345/345 [==============================] - 6s 16ms/step - loss: 2.2587 - accuracy: 0.2676\n",
      "Epoch 2/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 1.6624 - accuracy: 0.4540\n",
      "Epoch 3/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 1.1131 - accuracy: 0.6381\n",
      "Epoch 4/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.9123 - accuracy: 0.7143\n",
      "Epoch 5/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.7362 - accuracy: 0.7774\n",
      "Epoch 6/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.6626 - accuracy: 0.7977\n",
      "Epoch 7/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.6025 - accuracy: 0.8332\n",
      "Epoch 8/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.5084 - accuracy: 0.8434\n",
      "Epoch 9/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.4649 - accuracy: 0.8651\n",
      "Epoch 10/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.4486 - accuracy: 0.8687\n",
      "Epoch 11/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.4169 - accuracy: 0.8782\n",
      "Epoch 12/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.3897 - accuracy: 0.8832\n",
      "Epoch 13/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.3731 - accuracy: 0.8840\n",
      "Epoch 14/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.3419 - accuracy: 0.8912\n",
      "Epoch 15/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.3429 - accuracy: 0.8891\n",
      "Epoch 16/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.2951 - accuracy: 0.9072\n",
      "Epoch 17/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2877 - accuracy: 0.9065\n",
      "Epoch 18/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2931 - accuracy: 0.9173\n",
      "Epoch 19/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2538 - accuracy: 0.9275\n",
      "Epoch 20/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2595 - accuracy: 0.9195\n",
      "Epoch 21/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2368 - accuracy: 0.9347\n",
      "Epoch 22/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.3047 - accuracy: 0.9137\n",
      "Epoch 23/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2463 - accuracy: 0.9188\n",
      "Epoch 24/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2186 - accuracy: 0.9347\n",
      "Epoch 25/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2351 - accuracy: 0.9376\n",
      "Epoch 26/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2188 - accuracy: 0.9333\n",
      "Epoch 27/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2268 - accuracy: 0.9391\n",
      "Epoch 28/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1979 - accuracy: 0.9456\n",
      "Epoch 29/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2398 - accuracy: 0.9289\n",
      "Epoch 30/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1811 - accuracy: 0.9471\n",
      "Epoch 31/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2299 - accuracy: 0.9463\n",
      "Epoch 32/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2021 - accuracy: 0.9434\n",
      "Epoch 33/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1988 - accuracy: 0.9398\n",
      "Epoch 34/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1881 - accuracy: 0.9456\n",
      "Epoch 35/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.2109 - accuracy: 0.9478\n",
      "Epoch 36/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1667 - accuracy: 0.9463\n",
      "Epoch 37/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1550 - accuracy: 0.9543\n",
      "Epoch 38/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1859 - accuracy: 0.9485\n",
      "Epoch 39/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1455 - accuracy: 0.9543\n",
      "Epoch 40/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1829 - accuracy: 0.9536\n",
      "Epoch 41/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1656 - accuracy: 0.9485\n",
      "Epoch 42/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1474 - accuracy: 0.9558\n",
      "Epoch 43/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1556 - accuracy: 0.9572\n",
      "Epoch 44/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1578 - accuracy: 0.9608\n",
      "Epoch 45/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1368 - accuracy: 0.9616\n",
      "Epoch 46/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1374 - accuracy: 0.9594\n",
      "Epoch 47/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1666 - accuracy: 0.9507\n",
      "Epoch 48/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1566 - accuracy: 0.9572\n",
      "Epoch 49/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1490 - accuracy: 0.9543\n",
      "Epoch 50/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1251 - accuracy: 0.9608\n",
      "Epoch 51/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1358 - accuracy: 0.9543\n",
      "Epoch 52/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1156 - accuracy: 0.9616\n",
      "Epoch 53/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1369 - accuracy: 0.9616\n",
      "Epoch 54/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1085 - accuracy: 0.9659\n",
      "Epoch 55/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1403 - accuracy: 0.9630\n",
      "Epoch 56/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1010 - accuracy: 0.9637\n",
      "Epoch 57/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1305 - accuracy: 0.9630\n",
      "Epoch 58/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1114 - accuracy: 0.9695\n",
      "Epoch 59/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1204 - accuracy: 0.9630\n",
      "Epoch 60/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1268 - accuracy: 0.9616\n",
      "Epoch 61/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1217 - accuracy: 0.9703\n",
      "Epoch 62/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1055 - accuracy: 0.9674\n",
      "Epoch 63/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1329 - accuracy: 0.9630\n",
      "Epoch 64/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0915 - accuracy: 0.9717\n",
      "Epoch 65/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1249 - accuracy: 0.9637\n",
      "Epoch 66/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1218 - accuracy: 0.9637\n",
      "Epoch 67/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1248 - accuracy: 0.9674\n",
      "Epoch 68/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1230 - accuracy: 0.9688\n",
      "Epoch 69/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1060 - accuracy: 0.9674\n",
      "Epoch 70/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1341 - accuracy: 0.9623\n",
      "Epoch 71/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0862 - accuracy: 0.9717\n",
      "Epoch 72/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0860 - accuracy: 0.9768\n",
      "Epoch 73/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0867 - accuracy: 0.9724\n",
      "Epoch 74/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1156 - accuracy: 0.9666\n",
      "Epoch 75/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1312 - accuracy: 0.9608\n",
      "Epoch 76/500\n",
      "345/345 [==============================] - 6s 17ms/step - loss: 0.1280 - accuracy: 0.9630\n",
      "Epoch 77/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0733 - accuracy: 0.9768\n",
      "Epoch 78/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.1009 - accuracy: 0.9732\n",
      "Epoch 79/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1042 - accuracy: 0.9724\n",
      "Epoch 80/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0863 - accuracy: 0.9717\n",
      "Epoch 81/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.1034 - accuracy: 0.9717\n",
      "Epoch 82/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1043 - accuracy: 0.9652\n",
      "Epoch 83/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1158 - accuracy: 0.9645\n",
      "Epoch 84/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1170 - accuracy: 0.9732\n",
      "Epoch 85/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.1251 - accuracy: 0.9674\n",
      "Epoch 86/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.1296 - accuracy: 0.9710\n",
      "Epoch 87/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0903 - accuracy: 0.9710\n",
      "Epoch 88/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0887 - accuracy: 0.9732\n",
      "Epoch 89/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0966 - accuracy: 0.9710\n",
      "Epoch 90/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1066 - accuracy: 0.9688\n",
      "Epoch 91/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0800 - accuracy: 0.9848\n",
      "Epoch 92/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0731 - accuracy: 0.9797\n",
      "Epoch 93/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1063 - accuracy: 0.9739\n",
      "Epoch 94/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0705 - accuracy: 0.9768\n",
      "Epoch 95/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0896 - accuracy: 0.9710\n",
      "Epoch 96/500\n",
      "345/345 [==============================] - 6s 16ms/step - loss: 0.0764 - accuracy: 0.9753\n",
      "Epoch 97/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0894 - accuracy: 0.9753\n",
      "Epoch 98/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0810 - accuracy: 0.9761\n",
      "Epoch 99/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0983 - accuracy: 0.9732\n",
      "Epoch 100/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1125 - accuracy: 0.9688\n",
      "Epoch 101/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0883 - accuracy: 0.9761\n",
      "Epoch 102/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0799 - accuracy: 0.9775\n",
      "Epoch 103/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0833 - accuracy: 0.9768\n",
      "Epoch 104/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0931 - accuracy: 0.9739\n",
      "Epoch 105/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.1013 - accuracy: 0.9717\n",
      "Epoch 106/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0715 - accuracy: 0.9768\n",
      "Epoch 107/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0996 - accuracy: 0.9695\n",
      "Epoch 108/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0921 - accuracy: 0.9703\n",
      "Epoch 109/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0634 - accuracy: 0.9775\n",
      "Epoch 110/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0753 - accuracy: 0.9790\n",
      "Epoch 111/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0703 - accuracy: 0.9819\n",
      "Epoch 112/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0672 - accuracy: 0.9768\n",
      "Epoch 113/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0846 - accuracy: 0.9732\n",
      "Epoch 114/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0731 - accuracy: 0.9797\n",
      "Epoch 115/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0785 - accuracy: 0.9775\n",
      "Epoch 116/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0895 - accuracy: 0.9768\n",
      "Epoch 117/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0749 - accuracy: 0.9840\n",
      "Epoch 118/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0976 - accuracy: 0.9724\n",
      "Epoch 119/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0725 - accuracy: 0.9790\n",
      "Epoch 120/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0937 - accuracy: 0.9703\n",
      "Epoch 121/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0928 - accuracy: 0.9746\n",
      "Epoch 122/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0719 - accuracy: 0.9746\n",
      "Epoch 123/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0932 - accuracy: 0.9732\n",
      "Epoch 124/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0689 - accuracy: 0.9768\n",
      "Epoch 125/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0481 - accuracy: 0.9811\n",
      "Epoch 126/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0830 - accuracy: 0.9761\n",
      "Epoch 127/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0494 - accuracy: 0.9826\n",
      "Epoch 128/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0939 - accuracy: 0.9753\n",
      "Epoch 129/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0834 - accuracy: 0.9811\n",
      "Epoch 130/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0648 - accuracy: 0.9819\n",
      "Epoch 131/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0635 - accuracy: 0.9819\n",
      "Epoch 132/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0544 - accuracy: 0.9811\n",
      "Epoch 133/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0677 - accuracy: 0.9790\n",
      "Epoch 134/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0513 - accuracy: 0.9833\n",
      "Epoch 135/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1071 - accuracy: 0.9724\n",
      "Epoch 136/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0655 - accuracy: 0.9797\n",
      "Epoch 137/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0518 - accuracy: 0.9891\n",
      "Epoch 138/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0936 - accuracy: 0.9753\n",
      "Epoch 139/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0799 - accuracy: 0.9826\n",
      "Epoch 140/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0700 - accuracy: 0.9775\n",
      "Epoch 141/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0654 - accuracy: 0.9782\n",
      "Epoch 142/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0686 - accuracy: 0.9804\n",
      "Epoch 143/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0704 - accuracy: 0.9804\n",
      "Epoch 144/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0436 - accuracy: 0.9862\n",
      "Epoch 145/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0673 - accuracy: 0.9819\n",
      "Epoch 146/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0744 - accuracy: 0.9790\n",
      "Epoch 147/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0799 - accuracy: 0.9855\n",
      "Epoch 148/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0643 - accuracy: 0.9804\n",
      "Epoch 149/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0563 - accuracy: 0.9819\n",
      "Epoch 150/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0705 - accuracy: 0.9790\n",
      "Epoch 151/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0767 - accuracy: 0.9782\n",
      "Epoch 152/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0655 - accuracy: 0.9833\n",
      "Epoch 153/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0646 - accuracy: 0.9790\n",
      "Epoch 154/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0723 - accuracy: 0.9775\n",
      "Epoch 155/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0687 - accuracy: 0.9797\n",
      "Epoch 156/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0857 - accuracy: 0.9775\n",
      "Epoch 157/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0583 - accuracy: 0.9833\n",
      "Epoch 158/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0494 - accuracy: 0.9848\n",
      "Epoch 159/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0415 - accuracy: 0.9877\n",
      "Epoch 160/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0649 - accuracy: 0.9753\n",
      "Epoch 161/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0649 - accuracy: 0.9819\n",
      "Epoch 162/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0677 - accuracy: 0.9819\n",
      "Epoch 163/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0728 - accuracy: 0.9811\n",
      "Epoch 164/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0615 - accuracy: 0.9790\n",
      "Epoch 165/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0640 - accuracy: 0.9826\n",
      "Epoch 166/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0748 - accuracy: 0.9797\n",
      "Epoch 167/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0720 - accuracy: 0.9833\n",
      "Epoch 168/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.1208 - accuracy: 0.9739\n",
      "Epoch 169/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0476 - accuracy: 0.9855\n",
      "Epoch 170/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0697 - accuracy: 0.9826\n",
      "Epoch 171/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0495 - accuracy: 0.9833\n",
      "Epoch 172/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0356 - accuracy: 0.9869\n",
      "Epoch 173/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0575 - accuracy: 0.9898\n",
      "Epoch 174/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0519 - accuracy: 0.9826\n",
      "Epoch 175/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0521 - accuracy: 0.9855\n",
      "Epoch 176/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0938 - accuracy: 0.9826\n",
      "Epoch 177/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0509 - accuracy: 0.9869\n",
      "Epoch 178/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0698 - accuracy: 0.9819\n",
      "Epoch 179/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0612 - accuracy: 0.9826\n",
      "Epoch 180/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0514 - accuracy: 0.9819\n",
      "Epoch 181/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0383 - accuracy: 0.9891\n",
      "Epoch 182/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0850 - accuracy: 0.9761\n",
      "Epoch 183/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0738 - accuracy: 0.9819\n",
      "Epoch 184/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0581 - accuracy: 0.9862\n",
      "Epoch 185/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0522 - accuracy: 0.9855\n",
      "Epoch 186/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0631 - accuracy: 0.9833\n",
      "Epoch 187/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0558 - accuracy: 0.9862\n",
      "Epoch 188/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0495 - accuracy: 0.9840\n",
      "Epoch 189/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0219 - accuracy: 0.9920\n",
      "Epoch 190/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0579 - accuracy: 0.9891\n",
      "Epoch 191/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0602 - accuracy: 0.9840\n",
      "Epoch 192/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0489 - accuracy: 0.9891\n",
      "Epoch 193/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0250 - accuracy: 0.9913\n",
      "Epoch 194/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0433 - accuracy: 0.9855\n",
      "Epoch 195/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0456 - accuracy: 0.9840\n",
      "Epoch 196/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0595 - accuracy: 0.9848\n",
      "Epoch 197/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0638 - accuracy: 0.9826\n",
      "Epoch 198/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0873 - accuracy: 0.9775\n",
      "Epoch 199/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0844 - accuracy: 0.9797\n",
      "Epoch 200/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0586 - accuracy: 0.9811\n",
      "Epoch 201/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0820 - accuracy: 0.9782\n",
      "Epoch 202/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0643 - accuracy: 0.9833\n",
      "Epoch 203/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0667 - accuracy: 0.9797\n",
      "Epoch 204/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0506 - accuracy: 0.9877\n",
      "Epoch 205/500\n",
      "345/345 [==============================] - 5s 15ms/step - loss: 0.0694 - accuracy: 0.9768\n",
      "Epoch 206/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0667 - accuracy: 0.9804\n",
      "Epoch 207/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0432 - accuracy: 0.9898\n",
      "Epoch 208/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0524 - accuracy: 0.9855\n",
      "Epoch 209/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0358 - accuracy: 0.9848\n",
      "Epoch 210/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0412 - accuracy: 0.9884\n",
      "Epoch 211/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0535 - accuracy: 0.9833\n",
      "Epoch 212/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0504 - accuracy: 0.9826\n",
      "Epoch 213/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0634 - accuracy: 0.9811\n",
      "Epoch 214/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0492 - accuracy: 0.9869\n",
      "Epoch 215/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0605 - accuracy: 0.9819\n",
      "Epoch 216/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0540 - accuracy: 0.9884\n",
      "Epoch 217/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0478 - accuracy: 0.9869\n",
      "Epoch 218/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0548 - accuracy: 0.9804\n",
      "Epoch 219/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0749 - accuracy: 0.9775\n",
      "Epoch 220/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0743 - accuracy: 0.9811\n",
      "Epoch 221/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0570 - accuracy: 0.9855\n",
      "Epoch 222/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0583 - accuracy: 0.9826\n",
      "Epoch 223/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0478 - accuracy: 0.9862\n",
      "Epoch 224/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0741 - accuracy: 0.9833\n",
      "Epoch 225/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0460 - accuracy: 0.9826\n",
      "Epoch 226/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0331 - accuracy: 0.9862\n",
      "Epoch 227/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0543 - accuracy: 0.9855\n",
      "Epoch 228/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0341 - accuracy: 0.9898\n",
      "Epoch 229/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0430 - accuracy: 0.9862\n",
      "Epoch 230/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0719 - accuracy: 0.9804\n",
      "Epoch 231/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0770 - accuracy: 0.9848\n",
      "Epoch 232/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0498 - accuracy: 0.9855\n",
      "Epoch 233/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0591 - accuracy: 0.9782\n",
      "Epoch 234/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0463 - accuracy: 0.9848\n",
      "Epoch 235/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0504 - accuracy: 0.9848\n",
      "Epoch 236/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0587 - accuracy: 0.9848\n",
      "Epoch 237/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0615 - accuracy: 0.9848\n",
      "Epoch 238/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0484 - accuracy: 0.9877\n",
      "Epoch 239/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0481 - accuracy: 0.9877\n",
      "Epoch 240/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0553 - accuracy: 0.9869\n",
      "Epoch 241/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0776 - accuracy: 0.9790\n",
      "Epoch 242/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0368 - accuracy: 0.9877\n",
      "Epoch 243/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0857 - accuracy: 0.9811\n",
      "Epoch 244/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0612 - accuracy: 0.9840\n",
      "Epoch 245/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0400 - accuracy: 0.9898\n",
      "Epoch 246/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0391 - accuracy: 0.9906\n",
      "Epoch 247/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0814 - accuracy: 0.9848\n",
      "Epoch 248/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0732 - accuracy: 0.9826\n",
      "Epoch 249/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0519 - accuracy: 0.9826\n",
      "Epoch 250/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0559 - accuracy: 0.9833\n",
      "Epoch 251/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0589 - accuracy: 0.9869\n",
      "Epoch 252/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0576 - accuracy: 0.9811\n",
      "Epoch 253/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0632 - accuracy: 0.9848\n",
      "Epoch 254/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0356 - accuracy: 0.9898\n",
      "Epoch 255/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0310 - accuracy: 0.9913\n",
      "Epoch 256/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0559 - accuracy: 0.9884\n",
      "Epoch 257/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0810 - accuracy: 0.9790\n",
      "Epoch 258/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0618 - accuracy: 0.9840\n",
      "Epoch 259/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0390 - accuracy: 0.9920\n",
      "Epoch 260/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0354 - accuracy: 0.9891\n",
      "Epoch 261/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0477 - accuracy: 0.9869\n",
      "Epoch 262/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0404 - accuracy: 0.9869\n",
      "Epoch 263/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0425 - accuracy: 0.9913\n",
      "Epoch 264/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0792 - accuracy: 0.9804\n",
      "Epoch 265/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0593 - accuracy: 0.9826\n",
      "Epoch 266/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0505 - accuracy: 0.9891\n",
      "Epoch 267/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0663 - accuracy: 0.9826\n",
      "Epoch 268/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0548 - accuracy: 0.9877\n",
      "Epoch 269/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0324 - accuracy: 0.9862\n",
      "Epoch 270/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0437 - accuracy: 0.9884\n",
      "Epoch 271/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0599 - accuracy: 0.9862\n",
      "Epoch 272/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0423 - accuracy: 0.9862\n",
      "Epoch 273/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0651 - accuracy: 0.9848\n",
      "Epoch 274/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0573 - accuracy: 0.9848\n",
      "Epoch 275/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0605 - accuracy: 0.9869\n",
      "Epoch 276/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0304 - accuracy: 0.9913\n",
      "Epoch 277/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0367 - accuracy: 0.9891\n",
      "Epoch 278/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0369 - accuracy: 0.9913\n",
      "Epoch 279/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0544 - accuracy: 0.9848\n",
      "Epoch 280/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0333 - accuracy: 0.9884\n",
      "Epoch 281/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0589 - accuracy: 0.9819\n",
      "Epoch 282/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0391 - accuracy: 0.9913\n",
      "Epoch 283/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0678 - accuracy: 0.9869\n",
      "Epoch 284/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0513 - accuracy: 0.9862\n",
      "Epoch 285/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0411 - accuracy: 0.9869\n",
      "Epoch 286/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0563 - accuracy: 0.9819\n",
      "Epoch 287/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0379 - accuracy: 0.9877\n",
      "Epoch 288/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0427 - accuracy: 0.9848\n",
      "Epoch 289/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0746 - accuracy: 0.9884\n",
      "Epoch 290/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0598 - accuracy: 0.9877\n",
      "Epoch 291/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0594 - accuracy: 0.9877\n",
      "Epoch 292/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0508 - accuracy: 0.9877\n",
      "Epoch 293/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0444 - accuracy: 0.9869\n",
      "Epoch 294/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0340 - accuracy: 0.9913\n",
      "Epoch 295/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0303 - accuracy: 0.9927\n",
      "Epoch 296/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0508 - accuracy: 0.9869\n",
      "Epoch 297/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0616 - accuracy: 0.9811\n",
      "Epoch 298/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0751 - accuracy: 0.9790\n",
      "Epoch 299/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0544 - accuracy: 0.9797\n",
      "Epoch 300/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0521 - accuracy: 0.9826\n",
      "Epoch 301/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0377 - accuracy: 0.9906\n",
      "Epoch 302/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0369 - accuracy: 0.9913\n",
      "Epoch 303/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0368 - accuracy: 0.9906\n",
      "Epoch 304/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0715 - accuracy: 0.9833\n",
      "Epoch 305/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0370 - accuracy: 0.9840\n",
      "Epoch 306/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0426 - accuracy: 0.9891\n",
      "Epoch 307/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0449 - accuracy: 0.9869\n",
      "Epoch 308/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0694 - accuracy: 0.9935\n",
      "Epoch 309/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0277 - accuracy: 0.9906\n",
      "Epoch 310/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0451 - accuracy: 0.9869\n",
      "Epoch 311/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0449 - accuracy: 0.9862\n",
      "Epoch 312/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0514 - accuracy: 0.9862\n",
      "Epoch 313/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0251 - accuracy: 0.9935\n",
      "Epoch 314/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0474 - accuracy: 0.9884\n",
      "Epoch 315/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0222 - accuracy: 0.9949\n",
      "Epoch 316/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0435 - accuracy: 0.9891\n",
      "Epoch 317/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0464 - accuracy: 0.9898\n",
      "Epoch 318/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0566 - accuracy: 0.9891\n",
      "Epoch 319/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0369 - accuracy: 0.9869\n",
      "Epoch 320/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0373 - accuracy: 0.9913\n",
      "Epoch 321/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0727 - accuracy: 0.9848\n",
      "Epoch 322/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0527 - accuracy: 0.9877\n",
      "Epoch 323/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0502 - accuracy: 0.9855\n",
      "Epoch 324/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0682 - accuracy: 0.9848\n",
      "Epoch 325/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0374 - accuracy: 0.9891\n",
      "Epoch 326/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0595 - accuracy: 0.9869\n",
      "Epoch 327/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0449 - accuracy: 0.9869\n",
      "Epoch 328/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0596 - accuracy: 0.9840\n",
      "Epoch 329/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0501 - accuracy: 0.9862\n",
      "Epoch 330/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0471 - accuracy: 0.9833\n",
      "Epoch 331/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0402 - accuracy: 0.9877\n",
      "Epoch 332/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0313 - accuracy: 0.9884\n",
      "Epoch 333/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0428 - accuracy: 0.9884\n",
      "Epoch 334/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0545 - accuracy: 0.9906\n",
      "Epoch 335/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0428 - accuracy: 0.9898\n",
      "Epoch 336/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0398 - accuracy: 0.9906\n",
      "Epoch 337/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0511 - accuracy: 0.9891\n",
      "Epoch 338/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0591 - accuracy: 0.9848\n",
      "Epoch 339/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0614 - accuracy: 0.9840\n",
      "Epoch 340/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0460 - accuracy: 0.9920\n",
      "Epoch 341/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0394 - accuracy: 0.9869\n",
      "Epoch 342/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0384 - accuracy: 0.9884\n",
      "Epoch 343/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0785 - accuracy: 0.9826\n",
      "Epoch 344/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0477 - accuracy: 0.9898\n",
      "Epoch 345/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0410 - accuracy: 0.9877\n",
      "Epoch 346/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0566 - accuracy: 0.9913\n",
      "Epoch 347/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0240 - accuracy: 0.9949\n",
      "Epoch 348/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0483 - accuracy: 0.9906\n",
      "Epoch 349/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0694 - accuracy: 0.9833\n",
      "Epoch 350/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0597 - accuracy: 0.9855\n",
      "Epoch 351/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0414 - accuracy: 0.9898\n",
      "Epoch 352/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0459 - accuracy: 0.9877\n",
      "Epoch 353/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0360 - accuracy: 0.9884\n",
      "Epoch 354/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0451 - accuracy: 0.9848\n",
      "Epoch 355/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0499 - accuracy: 0.9855\n",
      "Epoch 356/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0388 - accuracy: 0.9906\n",
      "Epoch 357/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0530 - accuracy: 0.9898\n",
      "Epoch 358/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0581 - accuracy: 0.9862\n",
      "Epoch 359/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0409 - accuracy: 0.9855\n",
      "Epoch 360/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0514 - accuracy: 0.9906\n",
      "Epoch 361/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0604 - accuracy: 0.9869\n",
      "Epoch 362/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0501 - accuracy: 0.9862\n",
      "Epoch 363/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0581 - accuracy: 0.9848\n",
      "Epoch 364/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0421 - accuracy: 0.9833\n",
      "Epoch 365/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0429 - accuracy: 0.9884\n",
      "Epoch 366/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0494 - accuracy: 0.9840\n",
      "Epoch 367/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0387 - accuracy: 0.9862\n",
      "Epoch 368/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0400 - accuracy: 0.9906\n",
      "Epoch 369/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0612 - accuracy: 0.9898\n",
      "Epoch 370/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0354 - accuracy: 0.9869\n",
      "Epoch 371/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0548 - accuracy: 0.9855\n",
      "Epoch 372/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0484 - accuracy: 0.9848\n",
      "Epoch 373/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0286 - accuracy: 0.9898\n",
      "Epoch 374/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0433 - accuracy: 0.9877\n",
      "Epoch 375/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0243 - accuracy: 0.9920\n",
      "Epoch 376/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0343 - accuracy: 0.9906\n",
      "Epoch 377/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0470 - accuracy: 0.9833\n",
      "Epoch 378/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0390 - accuracy: 0.9898\n",
      "Epoch 379/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0291 - accuracy: 0.9891\n",
      "Epoch 380/500\n",
      "345/345 [==============================] - 5s 14ms/step - loss: 0.0982 - accuracy: 0.9862\n",
      "Epoch 381/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0584 - accuracy: 0.9826\n",
      "Epoch 382/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0480 - accuracy: 0.9884\n",
      "Epoch 383/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0501 - accuracy: 0.9884\n",
      "Epoch 384/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0619 - accuracy: 0.9855\n",
      "Epoch 385/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0244 - accuracy: 0.9927\n",
      "Epoch 386/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0527 - accuracy: 0.9877\n",
      "Epoch 387/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0684 - accuracy: 0.9833\n",
      "Epoch 388/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0304 - accuracy: 0.9891\n",
      "Epoch 389/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0476 - accuracy: 0.9877\n",
      "Epoch 390/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0688 - accuracy: 0.9869\n",
      "Epoch 391/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0428 - accuracy: 0.9862\n",
      "Epoch 392/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0323 - accuracy: 0.9920\n",
      "Epoch 393/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0374 - accuracy: 0.9913\n",
      "Epoch 394/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0479 - accuracy: 0.9906\n",
      "Epoch 395/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0425 - accuracy: 0.9884\n",
      "Epoch 396/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0283 - accuracy: 0.9913\n",
      "Epoch 397/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0345 - accuracy: 0.9891\n",
      "Epoch 398/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0500 - accuracy: 0.9855\n",
      "Epoch 399/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0418 - accuracy: 0.9898\n",
      "Epoch 400/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0260 - accuracy: 0.9927\n",
      "Epoch 401/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0414 - accuracy: 0.9920\n",
      "Epoch 402/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0496 - accuracy: 0.9913\n",
      "Epoch 403/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0603 - accuracy: 0.9862\n",
      "Epoch 404/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0551 - accuracy: 0.9869\n",
      "Epoch 405/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0447 - accuracy: 0.9913\n",
      "Epoch 406/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0531 - accuracy: 0.9869\n",
      "Epoch 407/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0558 - accuracy: 0.9877\n",
      "Epoch 408/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0438 - accuracy: 0.9906\n",
      "Epoch 409/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0380 - accuracy: 0.9877\n",
      "Epoch 410/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0451 - accuracy: 0.9927\n",
      "Epoch 411/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0216 - accuracy: 0.9927\n",
      "Epoch 412/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0228 - accuracy: 0.9920\n",
      "Epoch 413/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0721 - accuracy: 0.9862\n",
      "Epoch 414/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0600 - accuracy: 0.9891\n",
      "Epoch 415/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0401 - accuracy: 0.9898\n",
      "Epoch 416/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0429 - accuracy: 0.9884\n",
      "Epoch 417/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0374 - accuracy: 0.9891\n",
      "Epoch 418/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0248 - accuracy: 0.9913\n",
      "Epoch 419/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0264 - accuracy: 0.9906\n",
      "Epoch 420/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0773 - accuracy: 0.9804\n",
      "Epoch 421/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0392 - accuracy: 0.9891\n",
      "Epoch 422/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0453 - accuracy: 0.9898\n",
      "Epoch 423/500\n",
      "345/345 [==============================] - 4s 12ms/step - loss: 0.0418 - accuracy: 0.9855\n",
      "Epoch 424/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0367 - accuracy: 0.9906\n",
      "Epoch 425/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0413 - accuracy: 0.9898\n",
      "Epoch 426/500\n",
      "345/345 [==============================] - 5s 13ms/step - loss: 0.0634 - accuracy: 0.9877\n",
      "Epoch 427/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0246 - accuracy: 0.9920\n",
      "Epoch 428/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0326 - accuracy: 0.9920\n",
      "Epoch 429/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0350 - accuracy: 0.9877\n",
      "Epoch 430/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0374 - accuracy: 0.9877\n",
      "Epoch 431/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0493 - accuracy: 0.9884\n",
      "Epoch 432/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0426 - accuracy: 0.9891\n",
      "Epoch 433/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0344 - accuracy: 0.9935\n",
      "Epoch 434/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0473 - accuracy: 0.9891\n",
      "Epoch 435/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0584 - accuracy: 0.9884\n",
      "Epoch 436/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0469 - accuracy: 0.9906\n",
      "Epoch 437/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0295 - accuracy: 0.9913\n",
      "Epoch 438/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0438 - accuracy: 0.9898\n",
      "Epoch 439/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0395 - accuracy: 0.9913\n",
      "Epoch 440/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0452 - accuracy: 0.9891\n",
      "Epoch 441/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0375 - accuracy: 0.9906\n",
      "Epoch 442/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0380 - accuracy: 0.9920\n",
      "Epoch 443/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0517 - accuracy: 0.9840\n",
      "Epoch 444/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0439 - accuracy: 0.9898\n",
      "Epoch 445/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0395 - accuracy: 0.9877\n",
      "Epoch 446/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0421 - accuracy: 0.9869\n",
      "Epoch 447/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0416 - accuracy: 0.9884\n",
      "Epoch 448/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0268 - accuracy: 0.9942\n",
      "Epoch 449/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0396 - accuracy: 0.9884\n",
      "Epoch 450/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0680 - accuracy: 0.9840\n",
      "Epoch 451/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0608 - accuracy: 0.9855\n",
      "Epoch 452/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0504 - accuracy: 0.9913\n",
      "Epoch 453/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0352 - accuracy: 0.9891\n",
      "Epoch 454/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0314 - accuracy: 0.9906\n",
      "Epoch 455/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0230 - accuracy: 0.9927\n",
      "Epoch 456/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0459 - accuracy: 0.9891\n",
      "Epoch 457/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0344 - accuracy: 0.9884\n",
      "Epoch 458/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0632 - accuracy: 0.9869\n",
      "Epoch 459/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0311 - accuracy: 0.9927\n",
      "Epoch 460/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0499 - accuracy: 0.9898\n",
      "Epoch 461/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0686 - accuracy: 0.9855\n",
      "Epoch 462/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0527 - accuracy: 0.9877\n",
      "Epoch 463/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0168 - accuracy: 0.9942\n",
      "Epoch 464/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0298 - accuracy: 0.9935\n",
      "Epoch 465/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0466 - accuracy: 0.9869\n",
      "Epoch 466/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0384 - accuracy: 0.9935\n",
      "Epoch 467/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0321 - accuracy: 0.9920\n",
      "Epoch 468/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0462 - accuracy: 0.9869\n",
      "Epoch 469/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0362 - accuracy: 0.9913\n",
      "Epoch 470/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0364 - accuracy: 0.9927\n",
      "Epoch 471/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0224 - accuracy: 0.9913\n",
      "Epoch 472/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0195 - accuracy: 0.9978\n",
      "Epoch 473/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0651 - accuracy: 0.9855\n",
      "Epoch 474/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0331 - accuracy: 0.9891\n",
      "Epoch 475/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0609 - accuracy: 0.9877\n",
      "Epoch 476/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0417 - accuracy: 0.9906\n",
      "Epoch 477/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0382 - accuracy: 0.9927\n",
      "Epoch 478/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0395 - accuracy: 0.9927\n",
      "Epoch 479/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0332 - accuracy: 0.9927\n",
      "Epoch 480/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0367 - accuracy: 0.9884\n",
      "Epoch 481/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0277 - accuracy: 0.9906\n",
      "Epoch 482/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0381 - accuracy: 0.9906\n",
      "Epoch 483/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0330 - accuracy: 0.9956\n",
      "Epoch 484/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0673 - accuracy: 0.9833\n",
      "Epoch 485/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0403 - accuracy: 0.9942\n",
      "Epoch 486/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0552 - accuracy: 0.9855\n",
      "Epoch 487/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0249 - accuracy: 0.9898\n",
      "Epoch 488/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0204 - accuracy: 0.9906\n",
      "Epoch 489/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0577 - accuracy: 0.9862\n",
      "Epoch 490/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0462 - accuracy: 0.9869\n",
      "Epoch 491/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0247 - accuracy: 0.9891\n",
      "Epoch 492/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0381 - accuracy: 0.9898\n",
      "Epoch 493/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0337 - accuracy: 0.9913\n",
      "Epoch 494/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0269 - accuracy: 0.9935\n",
      "Epoch 495/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0278 - accuracy: 0.9935\n",
      "Epoch 496/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0301 - accuracy: 0.9913\n",
      "Epoch 497/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0622 - accuracy: 0.9898\n",
      "Epoch 498/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0512 - accuracy: 0.9891\n",
      "Epoch 499/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0301 - accuracy: 0.9927\n",
      "Epoch 500/500\n",
      "345/345 [==============================] - 4s 13ms/step - loss: 0.0362 - accuracy: 0.9913\n"
     ]
    }
   ],
   "source": [
    "Batch_Size = 4\n",
    "Shift_Range = 1\n",
    "Brightness_Range = 0.2\n",
    "Rotation_Angle = 5\n",
    "ZoomRange = 0.3\n",
    "\n",
    "datagen = ImageDataGenerator(width_shift_range=[-Shift_Range,Shift_Range], \n",
    "                             height_shift_range=[-Shift_Range,Shift_Range],\n",
    "                             brightness_range=[1-Brightness_Range,1+Brightness_Range],\n",
    "                             zoom_range=[1-ZoomRange, 1+ZoomRange],\n",
    "                             rotation_range=Rotation_Angle)\n",
    "\n",
    "if (Training_Percentage > 0):\n",
    "    train_iterator = datagen.flow(x_data, y_data, batch_size=Batch_Size)\n",
    "    validation_iterator = datagen.flow(X_test, y_test, batch_size=Batch_Size)\n",
    "    history = model.fit(train_iterator, validation_data = validation_iterator, epochs = Epoch_Anz)\n",
    "else:\n",
    "    train_iterator = datagen.flow(x_data, y_data, batch_size=Batch_Size)\n",
    "    history = model.fit(train_iterator, epochs = Epoch_Anz)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Learing result\n",
    " \n",
    "* Visualization of the training and validation results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_ges = np.append(loss_ges, history.history['loss'])\n",
    "plt.semilogy(history.history['loss'])\n",
    "\n",
    "if (Training_Percentage > 0):\n",
    "    val_loss_ges = np.append(val_loss_ges, history.history['val_loss'])\n",
    "    plt.semilogy(history.history['val_loss'])\n",
    "\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train','eval'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Check the model by hand\n",
    "\n",
    "* The following code uses the trained model to check the deviation for each picture.\n",
    "* x-axis walks through each pixel, y-axis shows the result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Check each image for expected and deviation\n",
    "* setting the switch \"only_deviation = true\" will only print the images for which the classification and the CNN-result deviates\n",
    "\n",
    "The output contains the following information:\n",
    "\n",
    "| Filename      | Expected Category           | Predicted Category        |\n",
    "|------------- |:-----------------------------:|--------------|\n",
    "| ziffer_sortiert_resize_NaN/5\\Ziffer_4_0034.jpg | 4  | -1 |\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-07-21 16:55:57.926647: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Input_dir='ziffer_sortiert_resize'\n",
    "res = []\n",
    "only_deviation = True\n",
    "show_wrong_image = True\n",
    "\n",
    "files = glob.glob(Input_dir + '/*.jpg')\n",
    "\n",
    "for aktfile in files:\n",
    "    base = os.path.basename(aktfile)\n",
    "    target = base[0:1]\n",
    "    if target == \"N\":\n",
    "        zw1 = -1\n",
    "    else:\n",
    "        zw1 = int(target)\n",
    "    expected_class = zw1\n",
    "    image_in = Image.open(aktfile)\n",
    "    test_image = np.array(image_in, dtype=\"float32\")\n",
    "    img = np.reshape(test_image,[1,32,20,3])\n",
    "    classes = np.argmax(model.predict(img, verbose=0), axis=-1)\n",
    "    classes = classes[0]\n",
    "    if classes == 10: \n",
    "        classes = -1\n",
    "    zw2 = classes\n",
    "    zw3 = zw2 - zw1\n",
    "    res.append(np.array([zw1, zw2, zw3]))\n",
    "    if only_deviation == True:\n",
    "        if str(classes) != str(expected_class):\n",
    "            print(aktfile + \" \" + str(expected_class) +  \" \" + str(classes))\n",
    "            if show_wrong_image == True:\n",
    "                display(image_in)\n",
    "    else:\n",
    "        print(aktfile + \" \" + aktsubdir +  \" \" + str(classes))\n",
    "        \n",
    "\n",
    "res = np.asarray(res)\n",
    "\n",
    "\n",
    "plt.plot(res[:,0])\n",
    "plt.plot(res[:,1])\n",
    "plt.title('Result')\n",
    "plt.ylabel('Digital Value')\n",
    "plt.xlabel('#Picture')\n",
    "plt.legend(['real','model'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Save the model\n",
    "\n",
    "* Save the model to the file with the \"h5\" file format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "export() takes 2 positional arguments but 3 were given",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[14], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[43mmodel\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mexport\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43mtest\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43mtf_saved_model\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m)\u001b[49m  \u001b[38;5;66;03m# replace tf.saved_model.save with this line\u001b[39;00m\n",
      "\u001b[0;31mTypeError\u001b[0m: export() takes 2 positional arguments but 3 were given"
     ]
    }
   ],
   "source": [
    "#model.export(\"test\", \"tf_saved_model\")  # replace tf.saved_model.save with this line"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dig-class11_1801_s2.tflite\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-07-21 17:22:44.265619: E tensorflow/core/framework/node_def_util.cc:676] NodeDef mentions attribute debug_name which is not in the op definition: Op<name=VarHandleOp; signature= -> resource:resource; attr=container:string,default=\"\"; attr=shared_name:string,default=\"\"; attr=dtype:type; attr=shape:shape; attr=allowed_devices:list(string),default=[]; is_stateful=true> This may be expected if your graph generating binary is newer  than this binary. Unknown attributes will be ignored. NodeDef: {{node batch_normalization/moving_mean}}\n",
      "2024-07-21 17:22:44.276177: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:364] Ignored output_format.\n",
      "2024-07-21 17:22:44.276186: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:367] Ignored drop_control_dependency.\n",
      "2024-07-21 17:22:44.276496: I tensorflow/cc/saved_model/reader.cc:45] Reading SavedModel from: test\n",
      "2024-07-21 17:22:44.277069: I tensorflow/cc/saved_model/reader.cc:91] Reading meta graph with tags { serve }\n",
      "2024-07-21 17:22:44.277073: I tensorflow/cc/saved_model/reader.cc:132] Reading SavedModel debug info (if present) from: test\n",
      "2024-07-21 17:22:44.278151: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:375] MLIR V1 optimization pass is not enabled\n",
      "2024-07-21 17:22:44.279102: I tensorflow/cc/saved_model/loader.cc:231] Restoring SavedModel bundle.\n",
      "2024-07-21 17:22:44.309888: E tensorflow/core/framework/node_def_util.cc:676] NodeDef mentions attribute debug_name which is not in the op definition: Op<name=VarHandleOp; signature= -> resource:resource; attr=container:string,default=\"\"; attr=shared_name:string,default=\"\"; attr=dtype:type; attr=shape:shape; attr=allowed_devices:list(string),default=[]; is_stateful=true> This may be expected if your graph generating binary is newer  than this binary. Unknown attributes will be ignored. NodeDef: {{node Variable}}\n",
      "2024-07-21 17:22:44.311639: I tensorflow/cc/saved_model/loader.cc:215] Running initialization op on SavedModel bundle at path: test\n",
      "2024-07-21 17:22:44.321600: I tensorflow/cc/saved_model/loader.cc:314] SavedModel load for tags { serve }; Status: success: OK. Took 45103 microseconds.\n",
      "2024-07-21 17:22:44.332106: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "357348"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FileName = TFlite_MainType + \"_\" + TFlite_Version + \"_\" + TFlite_Size + \".tflite\"\n",
    "\n",
    "converter = tf.lite.TFLiteConverter.from_saved_model(\"test\")\n",
    "tflite_model = converter.convert()\n",
    "open(FileName, \"wb\").write(tflite_model)\n",
    "print(FileName)\n",
    "Path(FileName).stat().st_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-07-21 17:22:47.808765: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:364] Ignored output_format.\n",
      "2024-07-21 17:22:47.808778: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:367] Ignored drop_control_dependency.\n",
      "2024-07-21 17:22:47.808930: I tensorflow/cc/saved_model/reader.cc:45] Reading SavedModel from: test\n",
      "2024-07-21 17:22:47.809442: I tensorflow/cc/saved_model/reader.cc:91] Reading meta graph with tags { serve }\n",
      "2024-07-21 17:22:47.809447: I tensorflow/cc/saved_model/reader.cc:132] Reading SavedModel debug info (if present) from: test\n",
      "2024-07-21 17:22:47.811249: I tensorflow/cc/saved_model/loader.cc:231] Restoring SavedModel bundle.\n",
      "2024-07-21 17:22:47.844302: I tensorflow/cc/saved_model/loader.cc:215] Running initialization op on SavedModel bundle at path: test\n",
      "2024-07-21 17:22:47.854692: I tensorflow/cc/saved_model/loader.cc:314] SavedModel load for tags { serve }; Status: success: OK. Took 45764 microseconds.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dig-class11_1801_s2_q.tflite\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "fully_quantize: 0, inference_type: 6, input_inference_type: FLOAT32, output_inference_type: FLOAT32\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "97856"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "FileName = TFlite_MainType + \"_\" + TFlite_Version + \"_\" + TFlite_Size + \"_q.tflite\"\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "def representative_dataset():\n",
    "    for n in range(x_data[0].size):\n",
    "      data = np.expand_dims(x_data[5], axis=0)\n",
    "      yield [data.astype(np.float32)]\n",
    "        \n",
    "converter2 = tf.lite.TFLiteConverter.from_saved_model(\"test\")\n",
    "converter2.representative_dataset = representative_dataset\n",
    "converter2.optimizations = [tf.lite.Optimize.DEFAULT]\n",
    "converter2.representative_dataset = representative_dataset\n",
    "tflite_quant_model = converter2.convert()\n",
    "\n",
    "open(FileName, \"wb\").write(tflite_quant_model)\n",
    "print(FileName)\n",
    "Path(FileName).stat().st_size"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Checking the images shows, that this are border line images, which can be interpreted as a good digit or a faulty one."
   ]
  }
 ],
 "metadata": {
  "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": "3.9.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}