{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "burmese_digits_recognition.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_Gg5PupVygkk",
        "outputId": "f7c20acd-a45f-436f-e215-1f4bdbfffe9f"
      },
      "source": [
        "# Install unrar package to work with rar files\n",
        "\n",
        "!pip install unrar==0.4"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting unrar==0.4\n",
            "  Downloading https://files.pythonhosted.org/packages/bb/0b/53130ccd483e3db8c8a460cb579bdb21b458d5494d67a261e1a5b273fbb9/unrar-0.4-py3-none-any.whl\n",
            "Installing collected packages: unrar\n",
            "Successfully installed unrar-0.4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PRi-4wGxEBB2"
      },
      "source": [
        "import os\n",
        "\n",
        "from glob import glob\n",
        "from multiprocessing import Pool, cpu_count\n",
        "\n",
        "import cv2\n",
        "import gdown\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from tensorflow import keras\n",
        "from google.colab.patches import cv2_imshow"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ymdz7ug_0Nrq",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "845e8ced-84d1-4822-b978-c37084b6bb87"
      },
      "source": [
        "# Download dataset from google drive\n",
        "# Source : https://www.kaggle.com/datasciencemlclub/burmesecharactersanddigit\n",
        "\n",
        "url = 'https://drive.google.com/uc?id=1JbD-xMQmcq6lT5gDjE5WiwJJ13ywe7J6'\n",
        "output = 'burmese_digits_and_characters.rar'\n",
        "\n",
        "if not os.path.exists(output):\n",
        "  gdown.download(url, output, quiet=False)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1JbD-xMQmcq6lT5gDjE5WiwJJ13ywe7J6\n",
            "To: /content/burmese_digits_and_characters.rar\n",
            "129MB [00:04, 27.7MB/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hN8daM4m1NUp"
      },
      "source": [
        "# Unrar the downloaded dataset\n",
        "\n",
        "source_folder = \"BurmeseCharacterDataSets\"\n",
        "\n",
        "if not os.path.exists(source_folder):\n",
        "  get_ipython().system_raw(\"unrar x '{}'\".format(output))"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "97asYMkQ4M3t"
      },
      "source": [
        "# Unrar the digits and store them in a folder named 'data'\n",
        "\n",
        "data_folder = \"data\"\n",
        "\n",
        "if not os.path.exists(data_folder):\n",
        "  get_ipython().system_raw(\"unrar x '{}/0.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/1.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/2.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/3.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/4.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/5.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/6.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/7.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/8.rar' '{}/'\".format(source_folder, data_folder))\n",
        "  get_ipython().system_raw(\"unrar x '{}/9.rar' '{}/'\".format(source_folder, data_folder))"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yCnal_O2-GKx",
        "outputId": "f133e6c8-1b1d-427a-96d4-f11c07395690"
      },
      "source": [
        "# Check the available number of images per class\n",
        "\n",
        "class_0 = glob(\"{}/0/*.jpg\".format(data_folder))\n",
        "class_1 = glob(\"{}/1/*.jpg\".format(data_folder))\n",
        "class_2 = glob(\"{}/2/*.jpg\".format(data_folder))\n",
        "class_3 = glob(\"{}/3/*.jpg\".format(data_folder))\n",
        "class_4 = glob(\"{}/4/*.jpg\".format(data_folder))\n",
        "class_5 = glob(\"{}/5/*.jpg\".format(data_folder))\n",
        "class_6 = glob(\"{}/6/*.jpg\".format(data_folder))\n",
        "class_7 = glob(\"{}/7/*.jpg\".format(data_folder))\n",
        "class_8 = glob(\"{}/8/*.jpg\".format(data_folder))\n",
        "class_9 = glob(\"{}/9/*.jpg\".format(data_folder))\n",
        "\n",
        "print(\"Number of images per class\")\n",
        "print(\"Class 0: {}\".format(len(class_0)))\n",
        "print(\"Class 1: {}\".format(len(class_1)))\n",
        "print(\"Class 2: {}\".format(len(class_2)))\n",
        "print(\"Class 3: {}\".format(len(class_3)))\n",
        "print(\"Class 4: {}\".format(len(class_4)))\n",
        "print(\"Class 5: {}\".format(len(class_5)))\n",
        "print(\"Class 6: {}\".format(len(class_6)))\n",
        "print(\"Class 7: {}\".format(len(class_7)))\n",
        "print(\"Class 8: {}\".format(len(class_8)))\n",
        "print(\"Class 9: {}\".format(len(class_9)))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of images per class\n",
            "Class 0: 222\n",
            "Class 1: 222\n",
            "Class 2: 222\n",
            "Class 3: 222\n",
            "Class 4: 222\n",
            "Class 5: 222\n",
            "Class 6: 222\n",
            "Class 7: 222\n",
            "Class 8: 222\n",
            "Class 9: 222\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 253
        },
        "id": "APZr1kF8oLdW",
        "outputId": "7b783dd5-e354-4c7f-8914-fabe41e3313d"
      },
      "source": [
        "# Let's take a look at one of the image\n",
        "\n",
        "img = cv2.imread(class_0[1])\n",
        "print(img.shape)\n",
        "cv2_imshow(img)"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(219, 310, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=310x219 at 0x7F6E4E5C97B8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 234
        },
        "id": "_4o_bf5Rkpip",
        "outputId": "74337596-2070-4eff-cda7-2d345416f31f"
      },
      "source": [
        "# It's quite big. Let's crop the image and make it square\n",
        "\n",
        "crop_img = img[10:210, 50:250]\n",
        "print(crop_img.shape)\n",
        "cv2_imshow(crop_img)"
      ],
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(200, 200, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=200x200 at 0x7F6E4E3979B0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 217
        },
        "id": "ORn8VcmyoSST",
        "outputId": "f17f778d-ee62-45ee-8eca-690a8e2de998"
      },
      "source": [
        "# Let's sharpen it to give more contrast\n",
        "\n",
        "sharpened_image = cv2.threshold(crop_img, 200, 255,cv2.THRESH_BINARY_INV)\n",
        "cv2_imshow(sharpened_image[-1])"
      ],
      "execution_count": 61,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=200x200 at 0x7F6E4E22A4A8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8fsV7DGrs-X2"
      },
      "source": [
        "# Preparing to store processed images\n",
        "new_path = \"data/processed/\"\n",
        "\n",
        "for i in range(10):\n",
        "  if not os.path.exists(new_path + str(i)):\n",
        "    os.makedirs(new_path + str(i))"
      ],
      "execution_count": 62,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A1t4a7mNjwk4"
      },
      "source": [
        "# Apply processing steps to all the images in each image folder\n",
        "\n",
        "def process_image(image):   \n",
        "  img = cv2.imread(image)\n",
        "  crop_img = img[10:210, 50:250]\n",
        "  _, sharpened_image = cv2.threshold(img, 200, 255,cv2.THRESH_BINARY_INV)\n",
        "\n",
        "  new_path = image.replace(\"data/\", \"data/processed/\")\n",
        "  cv2.imwrite(new_path, sharpened_image)\n",
        "\n",
        "\n",
        "if __name__ == \"__main__\":\n",
        "  with Pool(processes=cpu_count()-1) as pool:\n",
        "    pool.map(process_image, class_0)\n",
        "    pool.map(process_image, class_1)\n",
        "    pool.map(process_image, class_2)\n",
        "    pool.map(process_image, class_3)\n",
        "    pool.map(process_image, class_4)\n",
        "    pool.map(process_image, class_5)\n",
        "    pool.map(process_image, class_6)\n",
        "    pool.map(process_image, class_7)\n",
        "    pool.map(process_image, class_8)\n",
        "    pool.map(process_image, class_9)"
      ],
      "execution_count": 63,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ktO_3ZX34Vs8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7f8070f1-4963-452f-8656-0eb601ae0bd0"
      },
      "source": [
        "# Prepare images for model training: data augmentation, split into train and validation sets, etc\n",
        "\n",
        "batch_size = 32\n",
        "image_size=[200, 200]\n",
        "seed = 42\n",
        "val_split = 0.3\n",
        "\n",
        "np.seed(42)\n",
        "tf.random.set_seed(42)\n",
        "\n",
        "train_images = keras.preprocessing.image_dataset_from_directory(\n",
        "    new_path,\n",
        "    labels=\"inferred\",\n",
        "    label_mode=\"categorical\",\n",
        "    color_mode=\"rgb\",\n",
        "    batch_size=batch_size,\n",
        "    image_size=image_size,\n",
        "    shuffle=True,\n",
        "    seed=seed,\n",
        "    validation_split=val_split,\n",
        "    subset=\"training\",\n",
        "    interpolation=\"nearest\")\n",
        "\n",
        "val_images = keras.preprocessing.image_dataset_from_directory(\n",
        "    new_path,\n",
        "    labels=\"inferred\",\n",
        "    label_mode=\"categorical\",\n",
        "    color_mode=\"rgb\",\n",
        "    batch_size=batch_size,\n",
        "    image_size=image_size,\n",
        "    shuffle=True,\n",
        "    seed=seed,\n",
        "    validation_split=val_split,\n",
        "    subset=\"validation\",\n",
        "    interpolation=\"nearest\")\n",
        "\n",
        "# Prefect images so that I/O operations do not become a bottleneck\n",
        "train_images = train_images.prefetch(buffer_size=batch_size)\n",
        "val_images = val_images.prefetch(buffer_size=batch_size)"
      ],
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2220 files belonging to 10 classes.\n",
            "Using 1554 files for training.\n",
            "Found 2220 files belonging to 10 classes.\n",
            "Using 666 files for validation.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WOboNOCj5QjN",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 591
        },
        "outputId": "2679f415-f0fd-4c94-ac91-8110f21b8412"
      },
      "source": [
        "# Peek into the prepared images (sanity check)\n",
        "\n",
        "plt.figure(figsize=(10, 10))\n",
        "for images, labels in train_images.take(1):\n",
        "    for i in range(9):\n",
        "        ax = plt.subplot(3, 3, i + 1)\n",
        "        plt.imshow(images[i].numpy().astype(\"uint8\"))\n",
        "        plt.title(np.argmax(labels[i]))\n",
        "        plt.axis(\"off\")"
      ],
      "execution_count": 66,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 9 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 575
        },
        "id": "kYMW9d_HP9vl",
        "outputId": "5e66e785-9e06-4450-c52b-5c53b2819e76"
      },
      "source": [
        "# Augment train data by adding a few data augmentation tricks (provides more training data)\n",
        "\n",
        "data_augmentation = keras.Sequential(\n",
        "  [\n",
        "    keras.layers.experimental.preprocessing.RandomZoom(0.3),\n",
        "  ]\n",
        ")\n",
        "\n",
        "plt.figure(figsize=(10, 10))\n",
        "for images, _ in train_images.take(1):\n",
        "    for i in range(9):\n",
        "        augmented_images = data_augmentation(images)\n",
        "        ax = plt.subplot(3, 3, i + 1)\n",
        "        plt.imshow(augmented_images[0].numpy().astype(\"uint8\"))\n",
        "        plt.axis(\"off\")"
      ],
      "execution_count": 102,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 9 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8LokfRO7GgD3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3799442d-4d9a-4762-83f6-e69b7fef3399"
      },
      "source": [
        "# Download a pre-trained model (we are doing transfer learning)\n",
        "\n",
        "input_shape = image_size + [3]  # 3 channels (RGB)\n",
        "\n",
        "base_model = keras.applications.InceptionV3(weights='imagenet',  # load imaget net weights, \n",
        "                                         input_shape=input_shape, \n",
        "                                         include_top=False,  # don't include the top layer (classification layer)\n",
        "                                         pooling='avg')  # use average pooling method\n",
        "\n",
        "base_model.trainable = False\n",
        "base_model.summary()"
      ],
      "execution_count": 103,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"inception_v3\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_11 (InputLayer)           [(None, 200, 200, 3) 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_200 (Conv2D)             (None, 99, 99, 32)   864         input_11[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_200 (BatchN (None, 99, 99, 32)   96          conv2d_200[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_188 (Activation)     (None, 99, 99, 32)   0           batch_normalization_200[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_201 (Conv2D)             (None, 97, 97, 32)   9216        activation_188[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_201 (BatchN (None, 97, 97, 32)   96          conv2d_201[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_189 (Activation)     (None, 97, 97, 32)   0           batch_normalization_201[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_202 (Conv2D)             (None, 97, 97, 64)   18432       activation_189[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_202 (BatchN (None, 97, 97, 64)   192         conv2d_202[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_190 (Activation)     (None, 97, 97, 64)   0           batch_normalization_202[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_8 (MaxPooling2D)  (None, 48, 48, 64)   0           activation_190[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_203 (Conv2D)             (None, 48, 48, 80)   5120        max_pooling2d_8[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_203 (BatchN (None, 48, 48, 80)   240         conv2d_203[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_191 (Activation)     (None, 48, 48, 80)   0           batch_normalization_203[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_204 (Conv2D)             (None, 46, 46, 192)  138240      activation_191[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_204 (BatchN (None, 46, 46, 192)  576         conv2d_204[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_192 (Activation)     (None, 46, 46, 192)  0           batch_normalization_204[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_9 (MaxPooling2D)  (None, 22, 22, 192)  0           activation_192[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_208 (Conv2D)             (None, 22, 22, 64)   12288       max_pooling2d_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_208 (BatchN (None, 22, 22, 64)   192         conv2d_208[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_196 (Activation)     (None, 22, 22, 64)   0           batch_normalization_208[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_206 (Conv2D)             (None, 22, 22, 48)   9216        max_pooling2d_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_209 (Conv2D)             (None, 22, 22, 96)   55296       activation_196[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_206 (BatchN (None, 22, 22, 48)   144         conv2d_206[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_209 (BatchN (None, 22, 22, 96)   288         conv2d_209[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_194 (Activation)     (None, 22, 22, 48)   0           batch_normalization_206[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_197 (Activation)     (None, 22, 22, 96)   0           batch_normalization_209[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_18 (AveragePo (None, 22, 22, 192)  0           max_pooling2d_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_205 (Conv2D)             (None, 22, 22, 64)   12288       max_pooling2d_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_207 (Conv2D)             (None, 22, 22, 64)   76800       activation_194[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_210 (Conv2D)             (None, 22, 22, 96)   82944       activation_197[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_211 (Conv2D)             (None, 22, 22, 32)   6144        average_pooling2d_18[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_205 (BatchN (None, 22, 22, 64)   192         conv2d_205[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_207 (BatchN (None, 22, 22, 64)   192         conv2d_207[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_210 (BatchN (None, 22, 22, 96)   288         conv2d_210[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_211 (BatchN (None, 22, 22, 32)   96          conv2d_211[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_193 (Activation)     (None, 22, 22, 64)   0           batch_normalization_205[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_195 (Activation)     (None, 22, 22, 64)   0           batch_normalization_207[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_198 (Activation)     (None, 22, 22, 96)   0           batch_normalization_210[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_199 (Activation)     (None, 22, 22, 32)   0           batch_normalization_211[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed0 (Concatenate)            (None, 22, 22, 256)  0           activation_193[0][0]             \n",
            "                                                                 activation_195[0][0]             \n",
            "                                                                 activation_198[0][0]             \n",
            "                                                                 activation_199[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_215 (Conv2D)             (None, 22, 22, 64)   16384       mixed0[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_215 (BatchN (None, 22, 22, 64)   192         conv2d_215[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_203 (Activation)     (None, 22, 22, 64)   0           batch_normalization_215[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_213 (Conv2D)             (None, 22, 22, 48)   12288       mixed0[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_216 (Conv2D)             (None, 22, 22, 96)   55296       activation_203[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_213 (BatchN (None, 22, 22, 48)   144         conv2d_213[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_216 (BatchN (None, 22, 22, 96)   288         conv2d_216[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_201 (Activation)     (None, 22, 22, 48)   0           batch_normalization_213[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_204 (Activation)     (None, 22, 22, 96)   0           batch_normalization_216[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_19 (AveragePo (None, 22, 22, 256)  0           mixed0[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_212 (Conv2D)             (None, 22, 22, 64)   16384       mixed0[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_214 (Conv2D)             (None, 22, 22, 64)   76800       activation_201[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_217 (Conv2D)             (None, 22, 22, 96)   82944       activation_204[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_218 (Conv2D)             (None, 22, 22, 64)   16384       average_pooling2d_19[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_212 (BatchN (None, 22, 22, 64)   192         conv2d_212[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_214 (BatchN (None, 22, 22, 64)   192         conv2d_214[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_217 (BatchN (None, 22, 22, 96)   288         conv2d_217[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_218 (BatchN (None, 22, 22, 64)   192         conv2d_218[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_200 (Activation)     (None, 22, 22, 64)   0           batch_normalization_212[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_202 (Activation)     (None, 22, 22, 64)   0           batch_normalization_214[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_205 (Activation)     (None, 22, 22, 96)   0           batch_normalization_217[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_206 (Activation)     (None, 22, 22, 64)   0           batch_normalization_218[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed1 (Concatenate)            (None, 22, 22, 288)  0           activation_200[0][0]             \n",
            "                                                                 activation_202[0][0]             \n",
            "                                                                 activation_205[0][0]             \n",
            "                                                                 activation_206[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_222 (Conv2D)             (None, 22, 22, 64)   18432       mixed1[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_222 (BatchN (None, 22, 22, 64)   192         conv2d_222[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_210 (Activation)     (None, 22, 22, 64)   0           batch_normalization_222[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_220 (Conv2D)             (None, 22, 22, 48)   13824       mixed1[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_223 (Conv2D)             (None, 22, 22, 96)   55296       activation_210[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_220 (BatchN (None, 22, 22, 48)   144         conv2d_220[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_223 (BatchN (None, 22, 22, 96)   288         conv2d_223[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_208 (Activation)     (None, 22, 22, 48)   0           batch_normalization_220[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_211 (Activation)     (None, 22, 22, 96)   0           batch_normalization_223[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_20 (AveragePo (None, 22, 22, 288)  0           mixed1[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_219 (Conv2D)             (None, 22, 22, 64)   18432       mixed1[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_221 (Conv2D)             (None, 22, 22, 64)   76800       activation_208[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_224 (Conv2D)             (None, 22, 22, 96)   82944       activation_211[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_225 (Conv2D)             (None, 22, 22, 64)   18432       average_pooling2d_20[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_219 (BatchN (None, 22, 22, 64)   192         conv2d_219[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_221 (BatchN (None, 22, 22, 64)   192         conv2d_221[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_224 (BatchN (None, 22, 22, 96)   288         conv2d_224[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_225 (BatchN (None, 22, 22, 64)   192         conv2d_225[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_207 (Activation)     (None, 22, 22, 64)   0           batch_normalization_219[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_209 (Activation)     (None, 22, 22, 64)   0           batch_normalization_221[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_212 (Activation)     (None, 22, 22, 96)   0           batch_normalization_224[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_213 (Activation)     (None, 22, 22, 64)   0           batch_normalization_225[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed2 (Concatenate)            (None, 22, 22, 288)  0           activation_207[0][0]             \n",
            "                                                                 activation_209[0][0]             \n",
            "                                                                 activation_212[0][0]             \n",
            "                                                                 activation_213[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_227 (Conv2D)             (None, 22, 22, 64)   18432       mixed2[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_227 (BatchN (None, 22, 22, 64)   192         conv2d_227[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_215 (Activation)     (None, 22, 22, 64)   0           batch_normalization_227[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_228 (Conv2D)             (None, 22, 22, 96)   55296       activation_215[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_228 (BatchN (None, 22, 22, 96)   288         conv2d_228[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_216 (Activation)     (None, 22, 22, 96)   0           batch_normalization_228[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_226 (Conv2D)             (None, 10, 10, 384)  995328      mixed2[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_229 (Conv2D)             (None, 10, 10, 96)   82944       activation_216[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_226 (BatchN (None, 10, 10, 384)  1152        conv2d_226[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_229 (BatchN (None, 10, 10, 96)   288         conv2d_229[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_214 (Activation)     (None, 10, 10, 384)  0           batch_normalization_226[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_217 (Activation)     (None, 10, 10, 96)   0           batch_normalization_229[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_10 (MaxPooling2D) (None, 10, 10, 288)  0           mixed2[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "mixed3 (Concatenate)            (None, 10, 10, 768)  0           activation_214[0][0]             \n",
            "                                                                 activation_217[0][0]             \n",
            "                                                                 max_pooling2d_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_234 (Conv2D)             (None, 10, 10, 128)  98304       mixed3[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_234 (BatchN (None, 10, 10, 128)  384         conv2d_234[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_222 (Activation)     (None, 10, 10, 128)  0           batch_normalization_234[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_235 (Conv2D)             (None, 10, 10, 128)  114688      activation_222[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_235 (BatchN (None, 10, 10, 128)  384         conv2d_235[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_223 (Activation)     (None, 10, 10, 128)  0           batch_normalization_235[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_231 (Conv2D)             (None, 10, 10, 128)  98304       mixed3[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_236 (Conv2D)             (None, 10, 10, 128)  114688      activation_223[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_231 (BatchN (None, 10, 10, 128)  384         conv2d_231[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_236 (BatchN (None, 10, 10, 128)  384         conv2d_236[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_219 (Activation)     (None, 10, 10, 128)  0           batch_normalization_231[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_224 (Activation)     (None, 10, 10, 128)  0           batch_normalization_236[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_232 (Conv2D)             (None, 10, 10, 128)  114688      activation_219[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_237 (Conv2D)             (None, 10, 10, 128)  114688      activation_224[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_232 (BatchN (None, 10, 10, 128)  384         conv2d_232[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_237 (BatchN (None, 10, 10, 128)  384         conv2d_237[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_220 (Activation)     (None, 10, 10, 128)  0           batch_normalization_232[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_225 (Activation)     (None, 10, 10, 128)  0           batch_normalization_237[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_21 (AveragePo (None, 10, 10, 768)  0           mixed3[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_230 (Conv2D)             (None, 10, 10, 192)  147456      mixed3[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_233 (Conv2D)             (None, 10, 10, 192)  172032      activation_220[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_238 (Conv2D)             (None, 10, 10, 192)  172032      activation_225[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_239 (Conv2D)             (None, 10, 10, 192)  147456      average_pooling2d_21[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_230 (BatchN (None, 10, 10, 192)  576         conv2d_230[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_233 (BatchN (None, 10, 10, 192)  576         conv2d_233[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_238 (BatchN (None, 10, 10, 192)  576         conv2d_238[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_239 (BatchN (None, 10, 10, 192)  576         conv2d_239[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_218 (Activation)     (None, 10, 10, 192)  0           batch_normalization_230[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_221 (Activation)     (None, 10, 10, 192)  0           batch_normalization_233[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_226 (Activation)     (None, 10, 10, 192)  0           batch_normalization_238[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_227 (Activation)     (None, 10, 10, 192)  0           batch_normalization_239[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed4 (Concatenate)            (None, 10, 10, 768)  0           activation_218[0][0]             \n",
            "                                                                 activation_221[0][0]             \n",
            "                                                                 activation_226[0][0]             \n",
            "                                                                 activation_227[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_244 (Conv2D)             (None, 10, 10, 160)  122880      mixed4[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_244 (BatchN (None, 10, 10, 160)  480         conv2d_244[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_232 (Activation)     (None, 10, 10, 160)  0           batch_normalization_244[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_245 (Conv2D)             (None, 10, 10, 160)  179200      activation_232[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_245 (BatchN (None, 10, 10, 160)  480         conv2d_245[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_233 (Activation)     (None, 10, 10, 160)  0           batch_normalization_245[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_241 (Conv2D)             (None, 10, 10, 160)  122880      mixed4[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_246 (Conv2D)             (None, 10, 10, 160)  179200      activation_233[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_241 (BatchN (None, 10, 10, 160)  480         conv2d_241[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_246 (BatchN (None, 10, 10, 160)  480         conv2d_246[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_229 (Activation)     (None, 10, 10, 160)  0           batch_normalization_241[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_234 (Activation)     (None, 10, 10, 160)  0           batch_normalization_246[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_242 (Conv2D)             (None, 10, 10, 160)  179200      activation_229[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_247 (Conv2D)             (None, 10, 10, 160)  179200      activation_234[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_242 (BatchN (None, 10, 10, 160)  480         conv2d_242[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_247 (BatchN (None, 10, 10, 160)  480         conv2d_247[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_230 (Activation)     (None, 10, 10, 160)  0           batch_normalization_242[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_235 (Activation)     (None, 10, 10, 160)  0           batch_normalization_247[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_22 (AveragePo (None, 10, 10, 768)  0           mixed4[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_240 (Conv2D)             (None, 10, 10, 192)  147456      mixed4[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_243 (Conv2D)             (None, 10, 10, 192)  215040      activation_230[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_248 (Conv2D)             (None, 10, 10, 192)  215040      activation_235[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_249 (Conv2D)             (None, 10, 10, 192)  147456      average_pooling2d_22[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_240 (BatchN (None, 10, 10, 192)  576         conv2d_240[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_243 (BatchN (None, 10, 10, 192)  576         conv2d_243[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_248 (BatchN (None, 10, 10, 192)  576         conv2d_248[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_249 (BatchN (None, 10, 10, 192)  576         conv2d_249[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_228 (Activation)     (None, 10, 10, 192)  0           batch_normalization_240[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_231 (Activation)     (None, 10, 10, 192)  0           batch_normalization_243[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_236 (Activation)     (None, 10, 10, 192)  0           batch_normalization_248[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_237 (Activation)     (None, 10, 10, 192)  0           batch_normalization_249[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed5 (Concatenate)            (None, 10, 10, 768)  0           activation_228[0][0]             \n",
            "                                                                 activation_231[0][0]             \n",
            "                                                                 activation_236[0][0]             \n",
            "                                                                 activation_237[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_254 (Conv2D)             (None, 10, 10, 160)  122880      mixed5[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_254 (BatchN (None, 10, 10, 160)  480         conv2d_254[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_242 (Activation)     (None, 10, 10, 160)  0           batch_normalization_254[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_255 (Conv2D)             (None, 10, 10, 160)  179200      activation_242[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_255 (BatchN (None, 10, 10, 160)  480         conv2d_255[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_243 (Activation)     (None, 10, 10, 160)  0           batch_normalization_255[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_251 (Conv2D)             (None, 10, 10, 160)  122880      mixed5[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_256 (Conv2D)             (None, 10, 10, 160)  179200      activation_243[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_251 (BatchN (None, 10, 10, 160)  480         conv2d_251[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_256 (BatchN (None, 10, 10, 160)  480         conv2d_256[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_239 (Activation)     (None, 10, 10, 160)  0           batch_normalization_251[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_244 (Activation)     (None, 10, 10, 160)  0           batch_normalization_256[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_252 (Conv2D)             (None, 10, 10, 160)  179200      activation_239[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_257 (Conv2D)             (None, 10, 10, 160)  179200      activation_244[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_252 (BatchN (None, 10, 10, 160)  480         conv2d_252[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_257 (BatchN (None, 10, 10, 160)  480         conv2d_257[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_240 (Activation)     (None, 10, 10, 160)  0           batch_normalization_252[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_245 (Activation)     (None, 10, 10, 160)  0           batch_normalization_257[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_23 (AveragePo (None, 10, 10, 768)  0           mixed5[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_250 (Conv2D)             (None, 10, 10, 192)  147456      mixed5[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_253 (Conv2D)             (None, 10, 10, 192)  215040      activation_240[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_258 (Conv2D)             (None, 10, 10, 192)  215040      activation_245[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_259 (Conv2D)             (None, 10, 10, 192)  147456      average_pooling2d_23[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_250 (BatchN (None, 10, 10, 192)  576         conv2d_250[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_253 (BatchN (None, 10, 10, 192)  576         conv2d_253[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_258 (BatchN (None, 10, 10, 192)  576         conv2d_258[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_259 (BatchN (None, 10, 10, 192)  576         conv2d_259[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_238 (Activation)     (None, 10, 10, 192)  0           batch_normalization_250[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_241 (Activation)     (None, 10, 10, 192)  0           batch_normalization_253[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_246 (Activation)     (None, 10, 10, 192)  0           batch_normalization_258[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_247 (Activation)     (None, 10, 10, 192)  0           batch_normalization_259[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed6 (Concatenate)            (None, 10, 10, 768)  0           activation_238[0][0]             \n",
            "                                                                 activation_241[0][0]             \n",
            "                                                                 activation_246[0][0]             \n",
            "                                                                 activation_247[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_264 (Conv2D)             (None, 10, 10, 192)  147456      mixed6[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_264 (BatchN (None, 10, 10, 192)  576         conv2d_264[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_252 (Activation)     (None, 10, 10, 192)  0           batch_normalization_264[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_265 (Conv2D)             (None, 10, 10, 192)  258048      activation_252[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_265 (BatchN (None, 10, 10, 192)  576         conv2d_265[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_253 (Activation)     (None, 10, 10, 192)  0           batch_normalization_265[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_261 (Conv2D)             (None, 10, 10, 192)  147456      mixed6[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_266 (Conv2D)             (None, 10, 10, 192)  258048      activation_253[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_261 (BatchN (None, 10, 10, 192)  576         conv2d_261[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_266 (BatchN (None, 10, 10, 192)  576         conv2d_266[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_249 (Activation)     (None, 10, 10, 192)  0           batch_normalization_261[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_254 (Activation)     (None, 10, 10, 192)  0           batch_normalization_266[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_262 (Conv2D)             (None, 10, 10, 192)  258048      activation_249[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_267 (Conv2D)             (None, 10, 10, 192)  258048      activation_254[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_262 (BatchN (None, 10, 10, 192)  576         conv2d_262[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_267 (BatchN (None, 10, 10, 192)  576         conv2d_267[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_250 (Activation)     (None, 10, 10, 192)  0           batch_normalization_262[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_255 (Activation)     (None, 10, 10, 192)  0           batch_normalization_267[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_24 (AveragePo (None, 10, 10, 768)  0           mixed6[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_260 (Conv2D)             (None, 10, 10, 192)  147456      mixed6[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_263 (Conv2D)             (None, 10, 10, 192)  258048      activation_250[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_268 (Conv2D)             (None, 10, 10, 192)  258048      activation_255[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_269 (Conv2D)             (None, 10, 10, 192)  147456      average_pooling2d_24[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_260 (BatchN (None, 10, 10, 192)  576         conv2d_260[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_263 (BatchN (None, 10, 10, 192)  576         conv2d_263[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_268 (BatchN (None, 10, 10, 192)  576         conv2d_268[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_269 (BatchN (None, 10, 10, 192)  576         conv2d_269[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_248 (Activation)     (None, 10, 10, 192)  0           batch_normalization_260[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_251 (Activation)     (None, 10, 10, 192)  0           batch_normalization_263[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_256 (Activation)     (None, 10, 10, 192)  0           batch_normalization_268[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_257 (Activation)     (None, 10, 10, 192)  0           batch_normalization_269[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed7 (Concatenate)            (None, 10, 10, 768)  0           activation_248[0][0]             \n",
            "                                                                 activation_251[0][0]             \n",
            "                                                                 activation_256[0][0]             \n",
            "                                                                 activation_257[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_272 (Conv2D)             (None, 10, 10, 192)  147456      mixed7[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_272 (BatchN (None, 10, 10, 192)  576         conv2d_272[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_260 (Activation)     (None, 10, 10, 192)  0           batch_normalization_272[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_273 (Conv2D)             (None, 10, 10, 192)  258048      activation_260[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_273 (BatchN (None, 10, 10, 192)  576         conv2d_273[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_261 (Activation)     (None, 10, 10, 192)  0           batch_normalization_273[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_270 (Conv2D)             (None, 10, 10, 192)  147456      mixed7[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_274 (Conv2D)             (None, 10, 10, 192)  258048      activation_261[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_270 (BatchN (None, 10, 10, 192)  576         conv2d_270[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_274 (BatchN (None, 10, 10, 192)  576         conv2d_274[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_258 (Activation)     (None, 10, 10, 192)  0           batch_normalization_270[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_262 (Activation)     (None, 10, 10, 192)  0           batch_normalization_274[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_271 (Conv2D)             (None, 4, 4, 320)    552960      activation_258[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_275 (Conv2D)             (None, 4, 4, 192)    331776      activation_262[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_271 (BatchN (None, 4, 4, 320)    960         conv2d_271[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_275 (BatchN (None, 4, 4, 192)    576         conv2d_275[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_259 (Activation)     (None, 4, 4, 320)    0           batch_normalization_271[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_263 (Activation)     (None, 4, 4, 192)    0           batch_normalization_275[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_11 (MaxPooling2D) (None, 4, 4, 768)    0           mixed7[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "mixed8 (Concatenate)            (None, 4, 4, 1280)   0           activation_259[0][0]             \n",
            "                                                                 activation_263[0][0]             \n",
            "                                                                 max_pooling2d_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_280 (Conv2D)             (None, 4, 4, 448)    573440      mixed8[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_280 (BatchN (None, 4, 4, 448)    1344        conv2d_280[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_268 (Activation)     (None, 4, 4, 448)    0           batch_normalization_280[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_277 (Conv2D)             (None, 4, 4, 384)    491520      mixed8[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_281 (Conv2D)             (None, 4, 4, 384)    1548288     activation_268[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_277 (BatchN (None, 4, 4, 384)    1152        conv2d_277[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_281 (BatchN (None, 4, 4, 384)    1152        conv2d_281[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_265 (Activation)     (None, 4, 4, 384)    0           batch_normalization_277[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_269 (Activation)     (None, 4, 4, 384)    0           batch_normalization_281[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_278 (Conv2D)             (None, 4, 4, 384)    442368      activation_265[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_279 (Conv2D)             (None, 4, 4, 384)    442368      activation_265[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_282 (Conv2D)             (None, 4, 4, 384)    442368      activation_269[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_283 (Conv2D)             (None, 4, 4, 384)    442368      activation_269[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_25 (AveragePo (None, 4, 4, 1280)   0           mixed8[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_276 (Conv2D)             (None, 4, 4, 320)    409600      mixed8[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_278 (BatchN (None, 4, 4, 384)    1152        conv2d_278[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_279 (BatchN (None, 4, 4, 384)    1152        conv2d_279[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_282 (BatchN (None, 4, 4, 384)    1152        conv2d_282[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_283 (BatchN (None, 4, 4, 384)    1152        conv2d_283[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_284 (Conv2D)             (None, 4, 4, 192)    245760      average_pooling2d_25[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_276 (BatchN (None, 4, 4, 320)    960         conv2d_276[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_266 (Activation)     (None, 4, 4, 384)    0           batch_normalization_278[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_267 (Activation)     (None, 4, 4, 384)    0           batch_normalization_279[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_270 (Activation)     (None, 4, 4, 384)    0           batch_normalization_282[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_271 (Activation)     (None, 4, 4, 384)    0           batch_normalization_283[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_284 (BatchN (None, 4, 4, 192)    576         conv2d_284[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_264 (Activation)     (None, 4, 4, 320)    0           batch_normalization_276[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed9_0 (Concatenate)          (None, 4, 4, 768)    0           activation_266[0][0]             \n",
            "                                                                 activation_267[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_4 (Concatenate)     (None, 4, 4, 768)    0           activation_270[0][0]             \n",
            "                                                                 activation_271[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_272 (Activation)     (None, 4, 4, 192)    0           batch_normalization_284[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed9 (Concatenate)            (None, 4, 4, 2048)   0           activation_264[0][0]             \n",
            "                                                                 mixed9_0[0][0]                   \n",
            "                                                                 concatenate_4[0][0]              \n",
            "                                                                 activation_272[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_289 (Conv2D)             (None, 4, 4, 448)    917504      mixed9[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_289 (BatchN (None, 4, 4, 448)    1344        conv2d_289[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_277 (Activation)     (None, 4, 4, 448)    0           batch_normalization_289[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_286 (Conv2D)             (None, 4, 4, 384)    786432      mixed9[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_290 (Conv2D)             (None, 4, 4, 384)    1548288     activation_277[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_286 (BatchN (None, 4, 4, 384)    1152        conv2d_286[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_290 (BatchN (None, 4, 4, 384)    1152        conv2d_290[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_274 (Activation)     (None, 4, 4, 384)    0           batch_normalization_286[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_278 (Activation)     (None, 4, 4, 384)    0           batch_normalization_290[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_287 (Conv2D)             (None, 4, 4, 384)    442368      activation_274[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_288 (Conv2D)             (None, 4, 4, 384)    442368      activation_274[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_291 (Conv2D)             (None, 4, 4, 384)    442368      activation_278[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_292 (Conv2D)             (None, 4, 4, 384)    442368      activation_278[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_26 (AveragePo (None, 4, 4, 2048)   0           mixed9[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_285 (Conv2D)             (None, 4, 4, 320)    655360      mixed9[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_287 (BatchN (None, 4, 4, 384)    1152        conv2d_287[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_288 (BatchN (None, 4, 4, 384)    1152        conv2d_288[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_291 (BatchN (None, 4, 4, 384)    1152        conv2d_291[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_292 (BatchN (None, 4, 4, 384)    1152        conv2d_292[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_293 (Conv2D)             (None, 4, 4, 192)    393216      average_pooling2d_26[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_285 (BatchN (None, 4, 4, 320)    960         conv2d_285[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_275 (Activation)     (None, 4, 4, 384)    0           batch_normalization_287[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_276 (Activation)     (None, 4, 4, 384)    0           batch_normalization_288[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_279 (Activation)     (None, 4, 4, 384)    0           batch_normalization_291[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_280 (Activation)     (None, 4, 4, 384)    0           batch_normalization_292[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_293 (BatchN (None, 4, 4, 192)    576         conv2d_293[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_273 (Activation)     (None, 4, 4, 320)    0           batch_normalization_285[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed9_1 (Concatenate)          (None, 4, 4, 768)    0           activation_275[0][0]             \n",
            "                                                                 activation_276[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_5 (Concatenate)     (None, 4, 4, 768)    0           activation_279[0][0]             \n",
            "                                                                 activation_280[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_281 (Activation)     (None, 4, 4, 192)    0           batch_normalization_293[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed10 (Concatenate)           (None, 4, 4, 2048)   0           activation_273[0][0]             \n",
            "                                                                 mixed9_1[0][0]                   \n",
            "                                                                 concatenate_5[0][0]              \n",
            "                                                                 activation_281[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "global_average_pooling2d_5 (Glo (None, 2048)         0           mixed10[0][0]                    \n",
            "==================================================================================================\n",
            "Total params: 21,802,784\n",
            "Trainable params: 0\n",
            "Non-trainable params: 21,802,784\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pFjz21ysRmQY",
        "outputId": "74e9b864-f869-45b6-ba7e-ff754c15b3b1"
      },
      "source": [
        "# Build the final classification model\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(keras.Input(shape=input_shape))\n",
        "model.add(data_augmentation)\n",
        "model.add(base_model)\n",
        "model.add(keras.layers.Dense(512, activation='relu'))\n",
        "model.add(keras.layers.Dropout(0.2, seed=seed))\n",
        "model.add(keras.layers.Dense(10, activation='softmax'))\n",
        "model.summary()\n",
        "\n",
        "model.compile(\n",
        "    loss=keras.losses.CategoricalCrossentropy(label_smoothing=0.1),\n",
        "    optimizer=keras.optimizers.Adam(learning_rate=0.0001),\n",
        "    metrics=['accuracy']\n",
        ")"
      ],
      "execution_count": 109,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_28\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "sequential_25 (Sequential)   (None, 200, 200, 3)       0         \n",
            "_________________________________________________________________\n",
            "inception_v3 (Functional)    (None, 2048)              21802784  \n",
            "_________________________________________________________________\n",
            "dense_13 (Dense)             (None, 512)               1049088   \n",
            "_________________________________________________________________\n",
            "dropout_6 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_14 (Dense)             (None, 10)                5130      \n",
            "=================================================================\n",
            "Total params: 22,857,002\n",
            "Trainable params: 1,054,218\n",
            "Non-trainable params: 21,802,784\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OmbEefLdP8zT"
      },
      "source": [
        "# Save the best model with lowest validation loss using checkpoint callback\n",
        "\n",
        "checkpoint_path = os.path.join(\"checkpoint\", \"best_model\")\n",
        "model_checkpoint_callback = keras.callbacks.ModelCheckpoint(\n",
        "    filepath=checkpoint_path,\n",
        "    save_weights_only=True,\n",
        "    monitor='val_loss',\n",
        "    mode='min',\n",
        "    save_best_only=True\n",
        ")"
      ],
      "execution_count": 110,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-J7kL9ofZpcP",
        "outputId": "de4af0b9-6de5-4af5-efc9-47846ebfb65b"
      },
      "source": [
        "epochs = 500\n",
        "\n",
        "r = model.fit(train_images,\n",
        "              epochs=epochs,\n",
        "              validation_data=val_images,\n",
        "              callbacks=[model_checkpoint_callback])"
      ],
      "execution_count": 111,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/500\n",
            "49/49 [==============================] - 7s 134ms/step - loss: 15.8981 - accuracy: 0.1461 - val_loss: 4.7451 - val_accuracy: 0.3498\n",
            "Epoch 2/500\n",
            "49/49 [==============================] - 6s 124ms/step - loss: 8.0123 - accuracy: 0.2960 - val_loss: 4.5251 - val_accuracy: 0.4384\n",
            "Epoch 3/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 5.4318 - accuracy: 0.3938 - val_loss: 3.0746 - val_accuracy: 0.5255\n",
            "Epoch 4/500\n",
            "49/49 [==============================] - 6s 127ms/step - loss: 3.7822 - accuracy: 0.4318 - val_loss: 2.2934 - val_accuracy: 0.5646\n",
            "Epoch 5/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 2.5627 - accuracy: 0.4961 - val_loss: 1.7972 - val_accuracy: 0.5931\n",
            "Epoch 6/500\n",
            "49/49 [==============================] - 6s 130ms/step - loss: 1.9805 - accuracy: 0.5232 - val_loss: 1.6029 - val_accuracy: 0.6066\n",
            "Epoch 7/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 1.7541 - accuracy: 0.5515 - val_loss: 1.6155 - val_accuracy: 0.5495\n",
            "Epoch 8/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 1.6974 - accuracy: 0.5534 - val_loss: 1.4525 - val_accuracy: 0.6111\n",
            "Epoch 9/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.5434 - accuracy: 0.6004 - val_loss: 1.4458 - val_accuracy: 0.6216\n",
            "Epoch 10/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.4644 - accuracy: 0.6255 - val_loss: 1.3990 - val_accuracy: 0.6306\n",
            "Epoch 11/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.4354 - accuracy: 0.6229 - val_loss: 1.3741 - val_accuracy: 0.6411\n",
            "Epoch 12/500\n",
            "49/49 [==============================] - 6s 128ms/step - loss: 1.4490 - accuracy: 0.6287 - val_loss: 1.3339 - val_accuracy: 0.6547\n",
            "Epoch 13/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 1.3881 - accuracy: 0.6429 - val_loss: 1.3539 - val_accuracy: 0.6456\n",
            "Epoch 14/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.3160 - accuracy: 0.6795 - val_loss: 1.2950 - val_accuracy: 0.6592\n",
            "Epoch 15/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 1.2961 - accuracy: 0.6763 - val_loss: 1.3335 - val_accuracy: 0.6456\n",
            "Epoch 16/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 1.3001 - accuracy: 0.6879 - val_loss: 1.3325 - val_accuracy: 0.6637\n",
            "Epoch 17/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.3239 - accuracy: 0.6789 - val_loss: 1.2565 - val_accuracy: 0.7042\n",
            "Epoch 18/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 1.2629 - accuracy: 0.6988 - val_loss: 1.2657 - val_accuracy: 0.6847\n",
            "Epoch 19/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 1.2346 - accuracy: 0.7104 - val_loss: 1.2288 - val_accuracy: 0.7177\n",
            "Epoch 20/500\n",
            "49/49 [==============================] - 6s 128ms/step - loss: 1.2389 - accuracy: 0.7085 - val_loss: 1.2240 - val_accuracy: 0.7177\n",
            "Epoch 21/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.1760 - accuracy: 0.7368 - val_loss: 1.2417 - val_accuracy: 0.7117\n",
            "Epoch 22/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.1805 - accuracy: 0.7349 - val_loss: 1.1893 - val_accuracy: 0.7147\n",
            "Epoch 23/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.1602 - accuracy: 0.7420 - val_loss: 1.1976 - val_accuracy: 0.7222\n",
            "Epoch 24/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.1710 - accuracy: 0.7407 - val_loss: 1.2189 - val_accuracy: 0.7267\n",
            "Epoch 25/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.1551 - accuracy: 0.7413 - val_loss: 1.2133 - val_accuracy: 0.6847\n",
            "Epoch 26/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.1387 - accuracy: 0.7587 - val_loss: 1.2126 - val_accuracy: 0.7282\n",
            "Epoch 27/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 1.1258 - accuracy: 0.7529 - val_loss: 1.1570 - val_accuracy: 0.7402\n",
            "Epoch 28/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.1183 - accuracy: 0.7645 - val_loss: 1.1973 - val_accuracy: 0.7282\n",
            "Epoch 29/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.1213 - accuracy: 0.7606 - val_loss: 1.1785 - val_accuracy: 0.7402\n",
            "Epoch 30/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.0828 - accuracy: 0.7838 - val_loss: 1.1467 - val_accuracy: 0.7538\n",
            "Epoch 31/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.0778 - accuracy: 0.7748 - val_loss: 1.1624 - val_accuracy: 0.7327\n",
            "Epoch 32/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.0557 - accuracy: 0.7960 - val_loss: 1.1634 - val_accuracy: 0.7387\n",
            "Epoch 33/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.0835 - accuracy: 0.7825 - val_loss: 1.1813 - val_accuracy: 0.7297\n",
            "Epoch 34/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 1.0575 - accuracy: 0.7806 - val_loss: 1.1580 - val_accuracy: 0.7492\n",
            "Epoch 35/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 1.0863 - accuracy: 0.7741 - val_loss: 1.1056 - val_accuracy: 0.7688\n",
            "Epoch 36/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.0573 - accuracy: 0.7838 - val_loss: 1.1400 - val_accuracy: 0.7372\n",
            "Epoch 37/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 1.0244 - accuracy: 0.8069 - val_loss: 1.1243 - val_accuracy: 0.7613\n",
            "Epoch 38/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.0269 - accuracy: 0.7992 - val_loss: 1.1216 - val_accuracy: 0.7492\n",
            "Epoch 39/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 1.0097 - accuracy: 0.8160 - val_loss: 1.1279 - val_accuracy: 0.7523\n",
            "Epoch 40/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.0214 - accuracy: 0.8063 - val_loss: 1.1193 - val_accuracy: 0.7492\n",
            "Epoch 41/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 1.0201 - accuracy: 0.8108 - val_loss: 1.1087 - val_accuracy: 0.7523\n",
            "Epoch 42/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 1.0090 - accuracy: 0.8063 - val_loss: 1.0727 - val_accuracy: 0.7673\n",
            "Epoch 43/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9957 - accuracy: 0.8179 - val_loss: 1.1169 - val_accuracy: 0.7553\n",
            "Epoch 44/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 1.0175 - accuracy: 0.8179 - val_loss: 1.0824 - val_accuracy: 0.7748\n",
            "Epoch 45/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.9691 - accuracy: 0.8359 - val_loss: 1.0796 - val_accuracy: 0.7748\n",
            "Epoch 46/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.9806 - accuracy: 0.8275 - val_loss: 1.1100 - val_accuracy: 0.7688\n",
            "Epoch 47/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.9828 - accuracy: 0.8308 - val_loss: 1.0565 - val_accuracy: 0.7838\n",
            "Epoch 48/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9577 - accuracy: 0.8366 - val_loss: 1.0861 - val_accuracy: 0.7643\n",
            "Epoch 49/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9713 - accuracy: 0.8423 - val_loss: 1.1160 - val_accuracy: 0.7523\n",
            "Epoch 50/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.9491 - accuracy: 0.8501 - val_loss: 1.0545 - val_accuracy: 0.8018\n",
            "Epoch 51/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.9700 - accuracy: 0.8359 - val_loss: 1.0853 - val_accuracy: 0.7928\n",
            "Epoch 52/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.9454 - accuracy: 0.8514 - val_loss: 1.0587 - val_accuracy: 0.7883\n",
            "Epoch 53/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9446 - accuracy: 0.8430 - val_loss: 1.0842 - val_accuracy: 0.7943\n",
            "Epoch 54/500\n",
            "49/49 [==============================] - 6s 131ms/step - loss: 0.9258 - accuracy: 0.8584 - val_loss: 1.0525 - val_accuracy: 0.7868\n",
            "Epoch 55/500\n",
            "49/49 [==============================] - 6s 129ms/step - loss: 0.9261 - accuracy: 0.8546 - val_loss: 1.0487 - val_accuracy: 0.7928\n",
            "Epoch 56/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9478 - accuracy: 0.8507 - val_loss: 1.0578 - val_accuracy: 0.7808\n",
            "Epoch 57/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9324 - accuracy: 0.8552 - val_loss: 1.1068 - val_accuracy: 0.7793\n",
            "Epoch 58/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.9296 - accuracy: 0.8559 - val_loss: 1.1281 - val_accuracy: 0.7508\n",
            "Epoch 59/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.9292 - accuracy: 0.8616 - val_loss: 1.0267 - val_accuracy: 0.8078\n",
            "Epoch 60/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.9156 - accuracy: 0.8604 - val_loss: 1.0589 - val_accuracy: 0.8003\n",
            "Epoch 61/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8909 - accuracy: 0.8758 - val_loss: 1.0292 - val_accuracy: 0.8123\n",
            "Epoch 62/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.9262 - accuracy: 0.8610 - val_loss: 1.0784 - val_accuracy: 0.7823\n",
            "Epoch 63/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.8857 - accuracy: 0.8726 - val_loss: 1.1067 - val_accuracy: 0.7763\n",
            "Epoch 64/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.9064 - accuracy: 0.8681 - val_loss: 1.0742 - val_accuracy: 0.7898\n",
            "Epoch 65/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8940 - accuracy: 0.8616 - val_loss: 1.0299 - val_accuracy: 0.8018\n",
            "Epoch 66/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.8980 - accuracy: 0.8694 - val_loss: 1.0197 - val_accuracy: 0.8078\n",
            "Epoch 67/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.8958 - accuracy: 0.8745 - val_loss: 1.0564 - val_accuracy: 0.7853\n",
            "Epoch 68/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8943 - accuracy: 0.8764 - val_loss: 1.0538 - val_accuracy: 0.7988\n",
            "Epoch 69/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.9004 - accuracy: 0.8694 - val_loss: 1.0431 - val_accuracy: 0.7943\n",
            "Epoch 70/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8943 - accuracy: 0.8764 - val_loss: 1.0510 - val_accuracy: 0.8033\n",
            "Epoch 71/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8992 - accuracy: 0.8681 - val_loss: 1.0434 - val_accuracy: 0.7943\n",
            "Epoch 72/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.8808 - accuracy: 0.8810 - val_loss: 0.9970 - val_accuracy: 0.8243\n",
            "Epoch 73/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8776 - accuracy: 0.8803 - val_loss: 1.0470 - val_accuracy: 0.7973\n",
            "Epoch 74/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8730 - accuracy: 0.8880 - val_loss: 1.0153 - val_accuracy: 0.8228\n",
            "Epoch 75/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8763 - accuracy: 0.8777 - val_loss: 1.0581 - val_accuracy: 0.7913\n",
            "Epoch 76/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.8746 - accuracy: 0.8835 - val_loss: 1.0731 - val_accuracy: 0.7868\n",
            "Epoch 77/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8811 - accuracy: 0.8790 - val_loss: 1.0395 - val_accuracy: 0.8123\n",
            "Epoch 78/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8958 - accuracy: 0.8803 - val_loss: 1.0421 - val_accuracy: 0.7838\n",
            "Epoch 79/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8769 - accuracy: 0.8687 - val_loss: 1.1057 - val_accuracy: 0.7763\n",
            "Epoch 80/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.8720 - accuracy: 0.8829 - val_loss: 1.0393 - val_accuracy: 0.8033\n",
            "Epoch 81/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8757 - accuracy: 0.8816 - val_loss: 1.0261 - val_accuracy: 0.8018\n",
            "Epoch 82/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8641 - accuracy: 0.8874 - val_loss: 1.0075 - val_accuracy: 0.8033\n",
            "Epoch 83/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8848 - accuracy: 0.8764 - val_loss: 1.0323 - val_accuracy: 0.8033\n",
            "Epoch 84/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8646 - accuracy: 0.8912 - val_loss: 1.0136 - val_accuracy: 0.8048\n",
            "Epoch 85/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.8249 - accuracy: 0.9106 - val_loss: 0.9843 - val_accuracy: 0.8243\n",
            "Epoch 86/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.8534 - accuracy: 0.8990 - val_loss: 1.0287 - val_accuracy: 0.8033\n",
            "Epoch 87/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8641 - accuracy: 0.8893 - val_loss: 1.0580 - val_accuracy: 0.7883\n",
            "Epoch 88/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8474 - accuracy: 0.8996 - val_loss: 1.0353 - val_accuracy: 0.7958\n",
            "Epoch 89/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8390 - accuracy: 0.8938 - val_loss: 1.0127 - val_accuracy: 0.8018\n",
            "Epoch 90/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8304 - accuracy: 0.9054 - val_loss: 0.9985 - val_accuracy: 0.8258\n",
            "Epoch 91/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8394 - accuracy: 0.8983 - val_loss: 0.9932 - val_accuracy: 0.8363\n",
            "Epoch 92/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.8643 - accuracy: 0.8848 - val_loss: 1.0247 - val_accuracy: 0.8078\n",
            "Epoch 93/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8537 - accuracy: 0.8951 - val_loss: 0.9955 - val_accuracy: 0.8153\n",
            "Epoch 94/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8347 - accuracy: 0.9041 - val_loss: 1.0163 - val_accuracy: 0.8228\n",
            "Epoch 95/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8441 - accuracy: 0.8951 - val_loss: 1.0228 - val_accuracy: 0.8168\n",
            "Epoch 96/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8263 - accuracy: 0.9093 - val_loss: 1.0031 - val_accuracy: 0.8123\n",
            "Epoch 97/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8482 - accuracy: 0.8945 - val_loss: 1.0014 - val_accuracy: 0.8048\n",
            "Epoch 98/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.8257 - accuracy: 0.9060 - val_loss: 0.9890 - val_accuracy: 0.8198\n",
            "Epoch 99/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8420 - accuracy: 0.8835 - val_loss: 1.0223 - val_accuracy: 0.7943\n",
            "Epoch 100/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8183 - accuracy: 0.9112 - val_loss: 0.9975 - val_accuracy: 0.8243\n",
            "Epoch 101/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.8234 - accuracy: 0.9112 - val_loss: 1.0415 - val_accuracy: 0.8078\n",
            "Epoch 102/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.8309 - accuracy: 0.9048 - val_loss: 0.9990 - val_accuracy: 0.8273\n",
            "Epoch 103/500\n",
            "49/49 [==============================] - 6s 129ms/step - loss: 0.8193 - accuracy: 0.9112 - val_loss: 0.9744 - val_accuracy: 0.8333\n",
            "Epoch 104/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.8230 - accuracy: 0.9099 - val_loss: 0.9771 - val_accuracy: 0.8183\n",
            "Epoch 105/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8288 - accuracy: 0.9080 - val_loss: 1.0104 - val_accuracy: 0.8273\n",
            "Epoch 106/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.8166 - accuracy: 0.9125 - val_loss: 1.0115 - val_accuracy: 0.8078\n",
            "Epoch 107/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8368 - accuracy: 0.8983 - val_loss: 0.9763 - val_accuracy: 0.8348\n",
            "Epoch 108/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8249 - accuracy: 0.9093 - val_loss: 0.9976 - val_accuracy: 0.8213\n",
            "Epoch 109/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7999 - accuracy: 0.9228 - val_loss: 1.0060 - val_accuracy: 0.8183\n",
            "Epoch 110/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.8172 - accuracy: 0.9138 - val_loss: 0.9637 - val_accuracy: 0.8333\n",
            "Epoch 111/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8074 - accuracy: 0.9189 - val_loss: 0.9999 - val_accuracy: 0.8198\n",
            "Epoch 112/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8208 - accuracy: 0.9041 - val_loss: 1.0126 - val_accuracy: 0.8048\n",
            "Epoch 113/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.8018 - accuracy: 0.9273 - val_loss: 0.9588 - val_accuracy: 0.8228\n",
            "Epoch 114/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.8064 - accuracy: 0.9118 - val_loss: 0.9716 - val_accuracy: 0.8348\n",
            "Epoch 115/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8087 - accuracy: 0.9163 - val_loss: 0.9828 - val_accuracy: 0.8348\n",
            "Epoch 116/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8025 - accuracy: 0.9189 - val_loss: 0.9592 - val_accuracy: 0.8408\n",
            "Epoch 117/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7961 - accuracy: 0.9215 - val_loss: 0.9709 - val_accuracy: 0.8333\n",
            "Epoch 118/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7980 - accuracy: 0.9163 - val_loss: 1.0030 - val_accuracy: 0.8108\n",
            "Epoch 119/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8138 - accuracy: 0.9163 - val_loss: 0.9766 - val_accuracy: 0.8303\n",
            "Epoch 120/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8228 - accuracy: 0.9022 - val_loss: 0.9670 - val_accuracy: 0.8318\n",
            "Epoch 121/500\n",
            "49/49 [==============================] - 6s 124ms/step - loss: 0.7984 - accuracy: 0.9305 - val_loss: 0.9396 - val_accuracy: 0.8498\n",
            "Epoch 122/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7812 - accuracy: 0.9337 - val_loss: 0.9851 - val_accuracy: 0.8093\n",
            "Epoch 123/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.8004 - accuracy: 0.9215 - val_loss: 0.9795 - val_accuracy: 0.8333\n",
            "Epoch 124/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7952 - accuracy: 0.9221 - val_loss: 1.0090 - val_accuracy: 0.8153\n",
            "Epoch 125/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7915 - accuracy: 0.9247 - val_loss: 0.9745 - val_accuracy: 0.8213\n",
            "Epoch 126/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7953 - accuracy: 0.9273 - val_loss: 0.9571 - val_accuracy: 0.8498\n",
            "Epoch 127/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7907 - accuracy: 0.9247 - val_loss: 0.9498 - val_accuracy: 0.8393\n",
            "Epoch 128/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.7858 - accuracy: 0.9260 - val_loss: 0.9711 - val_accuracy: 0.8498\n",
            "Epoch 129/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.7718 - accuracy: 0.9324 - val_loss: 0.9513 - val_accuracy: 0.8529\n",
            "Epoch 130/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7728 - accuracy: 0.9376 - val_loss: 0.9523 - val_accuracy: 0.8529\n",
            "Epoch 131/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7945 - accuracy: 0.9189 - val_loss: 0.9635 - val_accuracy: 0.8378\n",
            "Epoch 132/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7888 - accuracy: 0.9221 - val_loss: 0.9545 - val_accuracy: 0.8378\n",
            "Epoch 133/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7674 - accuracy: 0.9395 - val_loss: 0.9771 - val_accuracy: 0.8288\n",
            "Epoch 134/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7742 - accuracy: 0.9324 - val_loss: 0.9779 - val_accuracy: 0.8303\n",
            "Epoch 135/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7907 - accuracy: 0.9254 - val_loss: 0.9585 - val_accuracy: 0.8393\n",
            "Epoch 136/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7658 - accuracy: 0.9376 - val_loss: 0.9513 - val_accuracy: 0.8514\n",
            "Epoch 137/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7686 - accuracy: 0.9344 - val_loss: 0.9856 - val_accuracy: 0.8288\n",
            "Epoch 138/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7712 - accuracy: 0.9299 - val_loss: 0.9637 - val_accuracy: 0.8348\n",
            "Epoch 139/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7835 - accuracy: 0.9318 - val_loss: 0.9699 - val_accuracy: 0.8183\n",
            "Epoch 140/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7772 - accuracy: 0.9376 - val_loss: 0.9844 - val_accuracy: 0.8273\n",
            "Epoch 141/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7803 - accuracy: 0.9260 - val_loss: 0.9463 - val_accuracy: 0.8393\n",
            "Epoch 142/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.7685 - accuracy: 0.9254 - val_loss: 0.9461 - val_accuracy: 0.8468\n",
            "Epoch 143/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7634 - accuracy: 0.9350 - val_loss: 0.9775 - val_accuracy: 0.8408\n",
            "Epoch 144/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7781 - accuracy: 0.9363 - val_loss: 0.9786 - val_accuracy: 0.8258\n",
            "Epoch 145/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7743 - accuracy: 0.9299 - val_loss: 0.9691 - val_accuracy: 0.8348\n",
            "Epoch 146/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7749 - accuracy: 0.9356 - val_loss: 1.0158 - val_accuracy: 0.8138\n",
            "Epoch 147/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7590 - accuracy: 0.9356 - val_loss: 0.9499 - val_accuracy: 0.8498\n",
            "Epoch 148/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7482 - accuracy: 0.9453 - val_loss: 0.9726 - val_accuracy: 0.8348\n",
            "Epoch 149/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7548 - accuracy: 0.9459 - val_loss: 0.9651 - val_accuracy: 0.8408\n",
            "Epoch 150/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.7574 - accuracy: 0.9472 - val_loss: 0.9468 - val_accuracy: 0.8559\n",
            "Epoch 151/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.7580 - accuracy: 0.9376 - val_loss: 0.9505 - val_accuracy: 0.8438\n",
            "Epoch 152/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7403 - accuracy: 0.9511 - val_loss: 0.9821 - val_accuracy: 0.8183\n",
            "Epoch 153/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7444 - accuracy: 0.9459 - val_loss: 0.9447 - val_accuracy: 0.8453\n",
            "Epoch 154/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7528 - accuracy: 0.9382 - val_loss: 0.9478 - val_accuracy: 0.8303\n",
            "Epoch 155/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.7559 - accuracy: 0.9408 - val_loss: 0.9378 - val_accuracy: 0.8348\n",
            "Epoch 156/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.7518 - accuracy: 0.9414 - val_loss: 0.9745 - val_accuracy: 0.8318\n",
            "Epoch 157/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7501 - accuracy: 0.9453 - val_loss: 0.9601 - val_accuracy: 0.8348\n",
            "Epoch 158/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7616 - accuracy: 0.9324 - val_loss: 0.9465 - val_accuracy: 0.8363\n",
            "Epoch 159/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7709 - accuracy: 0.9363 - val_loss: 0.9819 - val_accuracy: 0.8243\n",
            "Epoch 160/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.7610 - accuracy: 0.9344 - val_loss: 0.9336 - val_accuracy: 0.8514\n",
            "Epoch 161/500\n",
            "49/49 [==============================] - 6s 127ms/step - loss: 0.7596 - accuracy: 0.9389 - val_loss: 0.9248 - val_accuracy: 0.8619\n",
            "Epoch 162/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.7418 - accuracy: 0.9421 - val_loss: 0.9596 - val_accuracy: 0.8288\n",
            "Epoch 163/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7564 - accuracy: 0.9350 - val_loss: 0.9450 - val_accuracy: 0.8348\n",
            "Epoch 164/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7371 - accuracy: 0.9472 - val_loss: 0.9732 - val_accuracy: 0.8183\n",
            "Epoch 165/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7534 - accuracy: 0.9395 - val_loss: 0.9487 - val_accuracy: 0.8378\n",
            "Epoch 166/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7292 - accuracy: 0.9556 - val_loss: 0.9690 - val_accuracy: 0.8303\n",
            "Epoch 167/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7300 - accuracy: 0.9530 - val_loss: 0.9249 - val_accuracy: 0.8559\n",
            "Epoch 168/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7246 - accuracy: 0.9543 - val_loss: 0.9536 - val_accuracy: 0.8468\n",
            "Epoch 169/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7444 - accuracy: 0.9485 - val_loss: 0.9707 - val_accuracy: 0.8363\n",
            "Epoch 170/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7506 - accuracy: 0.9369 - val_loss: 0.9345 - val_accuracy: 0.8514\n",
            "Epoch 171/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7328 - accuracy: 0.9524 - val_loss: 0.9353 - val_accuracy: 0.8453\n",
            "Epoch 172/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7285 - accuracy: 0.9556 - val_loss: 0.9749 - val_accuracy: 0.8348\n",
            "Epoch 173/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.7300 - accuracy: 0.9498 - val_loss: 0.9184 - val_accuracy: 0.8574\n",
            "Epoch 174/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7362 - accuracy: 0.9524 - val_loss: 0.9591 - val_accuracy: 0.8453\n",
            "Epoch 175/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7460 - accuracy: 0.9421 - val_loss: 0.9454 - val_accuracy: 0.8438\n",
            "Epoch 176/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7493 - accuracy: 0.9350 - val_loss: 0.9389 - val_accuracy: 0.8468\n",
            "Epoch 177/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7425 - accuracy: 0.9492 - val_loss: 0.9605 - val_accuracy: 0.8273\n",
            "Epoch 178/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7355 - accuracy: 0.9472 - val_loss: 0.9354 - val_accuracy: 0.8423\n",
            "Epoch 179/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7290 - accuracy: 0.9530 - val_loss: 0.9499 - val_accuracy: 0.8348\n",
            "Epoch 180/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.7313 - accuracy: 0.9550 - val_loss: 0.9840 - val_accuracy: 0.8303\n",
            "Epoch 181/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7336 - accuracy: 0.9479 - val_loss: 0.9207 - val_accuracy: 0.8529\n",
            "Epoch 182/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7402 - accuracy: 0.9453 - val_loss: 0.9349 - val_accuracy: 0.8453\n",
            "Epoch 183/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7291 - accuracy: 0.9530 - val_loss: 0.9726 - val_accuracy: 0.8363\n",
            "Epoch 184/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7417 - accuracy: 0.9447 - val_loss: 0.9414 - val_accuracy: 0.8438\n",
            "Epoch 185/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.7182 - accuracy: 0.9492 - val_loss: 0.9328 - val_accuracy: 0.8438\n",
            "Epoch 186/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7330 - accuracy: 0.9485 - val_loss: 0.9590 - val_accuracy: 0.8438\n",
            "Epoch 187/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7264 - accuracy: 0.9569 - val_loss: 0.9348 - val_accuracy: 0.8589\n",
            "Epoch 188/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7416 - accuracy: 0.9453 - val_loss: 0.9493 - val_accuracy: 0.8408\n",
            "Epoch 189/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7189 - accuracy: 0.9601 - val_loss: 0.9548 - val_accuracy: 0.8348\n",
            "Epoch 190/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7226 - accuracy: 0.9562 - val_loss: 0.9494 - val_accuracy: 0.8453\n",
            "Epoch 191/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.7294 - accuracy: 0.9492 - val_loss: 0.9452 - val_accuracy: 0.8529\n",
            "Epoch 192/500\n",
            "49/49 [==============================] - 5s 112ms/step - loss: 0.7287 - accuracy: 0.9543 - val_loss: 0.9303 - val_accuracy: 0.8498\n",
            "Epoch 193/500\n",
            "49/49 [==============================] - 6s 115ms/step - loss: 0.7315 - accuracy: 0.9569 - val_loss: 0.9468 - val_accuracy: 0.8303\n",
            "Epoch 194/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7208 - accuracy: 0.9556 - val_loss: 0.9435 - val_accuracy: 0.8514\n",
            "Epoch 195/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7221 - accuracy: 0.9537 - val_loss: 0.9344 - val_accuracy: 0.8408\n",
            "Epoch 196/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7387 - accuracy: 0.9408 - val_loss: 0.9455 - val_accuracy: 0.8423\n",
            "Epoch 197/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7159 - accuracy: 0.9588 - val_loss: 0.9394 - val_accuracy: 0.8589\n",
            "Epoch 198/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7167 - accuracy: 0.9530 - val_loss: 0.9382 - val_accuracy: 0.8498\n",
            "Epoch 199/500\n",
            "49/49 [==============================] - 6s 131ms/step - loss: 0.7297 - accuracy: 0.9447 - val_loss: 0.9134 - val_accuracy: 0.8724\n",
            "Epoch 200/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.7212 - accuracy: 0.9601 - val_loss: 0.9158 - val_accuracy: 0.8589\n",
            "Epoch 201/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7095 - accuracy: 0.9595 - val_loss: 0.9280 - val_accuracy: 0.8483\n",
            "Epoch 202/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7233 - accuracy: 0.9511 - val_loss: 0.9355 - val_accuracy: 0.8589\n",
            "Epoch 203/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7254 - accuracy: 0.9588 - val_loss: 0.9222 - val_accuracy: 0.8544\n",
            "Epoch 204/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7111 - accuracy: 0.9646 - val_loss: 0.9420 - val_accuracy: 0.8378\n",
            "Epoch 205/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7329 - accuracy: 0.9485 - val_loss: 0.9198 - val_accuracy: 0.8574\n",
            "Epoch 206/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7258 - accuracy: 0.9492 - val_loss: 0.9341 - val_accuracy: 0.8498\n",
            "Epoch 207/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7375 - accuracy: 0.9453 - val_loss: 0.9633 - val_accuracy: 0.8363\n",
            "Epoch 208/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.7175 - accuracy: 0.9524 - val_loss: 0.8998 - val_accuracy: 0.8634\n",
            "Epoch 209/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.7085 - accuracy: 0.9562 - val_loss: 0.9425 - val_accuracy: 0.8423\n",
            "Epoch 210/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.7209 - accuracy: 0.9498 - val_loss: 0.9287 - val_accuracy: 0.8619\n",
            "Epoch 211/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7093 - accuracy: 0.9627 - val_loss: 0.9032 - val_accuracy: 0.8709\n",
            "Epoch 212/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6975 - accuracy: 0.9691 - val_loss: 0.9301 - val_accuracy: 0.8559\n",
            "Epoch 213/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7253 - accuracy: 0.9517 - val_loss: 0.9276 - val_accuracy: 0.8544\n",
            "Epoch 214/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7051 - accuracy: 0.9627 - val_loss: 0.9247 - val_accuracy: 0.8559\n",
            "Epoch 215/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7168 - accuracy: 0.9601 - val_loss: 0.9097 - val_accuracy: 0.8589\n",
            "Epoch 216/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7099 - accuracy: 0.9614 - val_loss: 0.9294 - val_accuracy: 0.8498\n",
            "Epoch 217/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7067 - accuracy: 0.9607 - val_loss: 0.9407 - val_accuracy: 0.8453\n",
            "Epoch 218/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7041 - accuracy: 0.9653 - val_loss: 0.9235 - val_accuracy: 0.8574\n",
            "Epoch 219/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7221 - accuracy: 0.9505 - val_loss: 0.9328 - val_accuracy: 0.8483\n",
            "Epoch 220/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.7126 - accuracy: 0.9556 - val_loss: 0.9021 - val_accuracy: 0.8694\n",
            "Epoch 221/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6952 - accuracy: 0.9607 - val_loss: 0.9019 - val_accuracy: 0.8619\n",
            "Epoch 222/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6951 - accuracy: 0.9698 - val_loss: 0.9052 - val_accuracy: 0.8529\n",
            "Epoch 223/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7049 - accuracy: 0.9627 - val_loss: 0.9079 - val_accuracy: 0.8619\n",
            "Epoch 224/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6996 - accuracy: 0.9640 - val_loss: 0.9020 - val_accuracy: 0.8589\n",
            "Epoch 225/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7141 - accuracy: 0.9607 - val_loss: 0.9222 - val_accuracy: 0.8498\n",
            "Epoch 226/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.7005 - accuracy: 0.9672 - val_loss: 0.9381 - val_accuracy: 0.8378\n",
            "Epoch 227/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7262 - accuracy: 0.9479 - val_loss: 0.9067 - val_accuracy: 0.8544\n",
            "Epoch 228/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7124 - accuracy: 0.9614 - val_loss: 0.9324 - val_accuracy: 0.8498\n",
            "Epoch 229/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.7047 - accuracy: 0.9575 - val_loss: 0.9562 - val_accuracy: 0.8333\n",
            "Epoch 230/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7094 - accuracy: 0.9575 - val_loss: 0.9241 - val_accuracy: 0.8574\n",
            "Epoch 231/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7026 - accuracy: 0.9620 - val_loss: 0.9428 - val_accuracy: 0.8423\n",
            "Epoch 232/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6973 - accuracy: 0.9672 - val_loss: 0.9020 - val_accuracy: 0.8544\n",
            "Epoch 233/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6905 - accuracy: 0.9685 - val_loss: 0.9337 - val_accuracy: 0.8514\n",
            "Epoch 234/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7089 - accuracy: 0.9575 - val_loss: 0.9108 - val_accuracy: 0.8709\n",
            "Epoch 235/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7026 - accuracy: 0.9607 - val_loss: 0.9307 - val_accuracy: 0.8303\n",
            "Epoch 236/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.7061 - accuracy: 0.9607 - val_loss: 0.9383 - val_accuracy: 0.8468\n",
            "Epoch 237/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6842 - accuracy: 0.9717 - val_loss: 0.9222 - val_accuracy: 0.8529\n",
            "Epoch 238/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6977 - accuracy: 0.9633 - val_loss: 0.9188 - val_accuracy: 0.8514\n",
            "Epoch 239/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6998 - accuracy: 0.9646 - val_loss: 0.9178 - val_accuracy: 0.8574\n",
            "Epoch 240/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6950 - accuracy: 0.9646 - val_loss: 0.9357 - val_accuracy: 0.8529\n",
            "Epoch 241/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6937 - accuracy: 0.9698 - val_loss: 0.9108 - val_accuracy: 0.8589\n",
            "Epoch 242/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6960 - accuracy: 0.9633 - val_loss: 0.9087 - val_accuracy: 0.8483\n",
            "Epoch 243/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.7026 - accuracy: 0.9640 - val_loss: 0.9120 - val_accuracy: 0.8574\n",
            "Epoch 244/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6911 - accuracy: 0.9665 - val_loss: 0.9020 - val_accuracy: 0.8649\n",
            "Epoch 245/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6953 - accuracy: 0.9653 - val_loss: 0.9143 - val_accuracy: 0.8559\n",
            "Epoch 246/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6814 - accuracy: 0.9659 - val_loss: 0.9327 - val_accuracy: 0.8408\n",
            "Epoch 247/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6892 - accuracy: 0.9698 - val_loss: 0.9056 - val_accuracy: 0.8559\n",
            "Epoch 248/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6758 - accuracy: 0.9710 - val_loss: 0.9245 - val_accuracy: 0.8408\n",
            "Epoch 249/500\n",
            "49/49 [==============================] - 6s 129ms/step - loss: 0.6973 - accuracy: 0.9569 - val_loss: 0.8939 - val_accuracy: 0.8679\n",
            "Epoch 250/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6870 - accuracy: 0.9665 - val_loss: 0.9108 - val_accuracy: 0.8634\n",
            "Epoch 251/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6881 - accuracy: 0.9698 - val_loss: 0.9266 - val_accuracy: 0.8544\n",
            "Epoch 252/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6899 - accuracy: 0.9678 - val_loss: 0.9059 - val_accuracy: 0.8694\n",
            "Epoch 253/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6904 - accuracy: 0.9640 - val_loss: 0.9131 - val_accuracy: 0.8529\n",
            "Epoch 254/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6765 - accuracy: 0.9730 - val_loss: 0.9072 - val_accuracy: 0.8634\n",
            "Epoch 255/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6826 - accuracy: 0.9691 - val_loss: 0.8975 - val_accuracy: 0.8574\n",
            "Epoch 256/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6920 - accuracy: 0.9659 - val_loss: 0.9022 - val_accuracy: 0.8664\n",
            "Epoch 257/500\n",
            "49/49 [==============================] - 6s 128ms/step - loss: 0.6902 - accuracy: 0.9691 - val_loss: 0.8879 - val_accuracy: 0.8679\n",
            "Epoch 258/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6737 - accuracy: 0.9801 - val_loss: 0.9233 - val_accuracy: 0.8498\n",
            "Epoch 259/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6823 - accuracy: 0.9723 - val_loss: 0.8980 - val_accuracy: 0.8529\n",
            "Epoch 260/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6771 - accuracy: 0.9775 - val_loss: 0.9049 - val_accuracy: 0.8468\n",
            "Epoch 261/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6948 - accuracy: 0.9582 - val_loss: 0.8989 - val_accuracy: 0.8529\n",
            "Epoch 262/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6932 - accuracy: 0.9627 - val_loss: 0.8940 - val_accuracy: 0.8754\n",
            "Epoch 263/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6851 - accuracy: 0.9704 - val_loss: 0.9152 - val_accuracy: 0.8468\n",
            "Epoch 264/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6832 - accuracy: 0.9685 - val_loss: 0.9150 - val_accuracy: 0.8529\n",
            "Epoch 265/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6847 - accuracy: 0.9672 - val_loss: 0.9038 - val_accuracy: 0.8589\n",
            "Epoch 266/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6844 - accuracy: 0.9659 - val_loss: 0.9224 - val_accuracy: 0.8649\n",
            "Epoch 267/500\n",
            "49/49 [==============================] - 6s 124ms/step - loss: 0.6946 - accuracy: 0.9582 - val_loss: 0.8988 - val_accuracy: 0.8574\n",
            "Epoch 268/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6927 - accuracy: 0.9607 - val_loss: 0.8977 - val_accuracy: 0.8619\n",
            "Epoch 269/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6751 - accuracy: 0.9749 - val_loss: 0.9151 - val_accuracy: 0.8498\n",
            "Epoch 270/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.7000 - accuracy: 0.9595 - val_loss: 0.9139 - val_accuracy: 0.8468\n",
            "Epoch 271/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6824 - accuracy: 0.9698 - val_loss: 0.9298 - val_accuracy: 0.8483\n",
            "Epoch 272/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6738 - accuracy: 0.9794 - val_loss: 0.9076 - val_accuracy: 0.8589\n",
            "Epoch 273/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6883 - accuracy: 0.9633 - val_loss: 0.8952 - val_accuracy: 0.8544\n",
            "Epoch 274/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6824 - accuracy: 0.9736 - val_loss: 0.8973 - val_accuracy: 0.8709\n",
            "Epoch 275/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6862 - accuracy: 0.9717 - val_loss: 0.9221 - val_accuracy: 0.8483\n",
            "Epoch 276/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6799 - accuracy: 0.9743 - val_loss: 0.9375 - val_accuracy: 0.8498\n",
            "Epoch 277/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6735 - accuracy: 0.9768 - val_loss: 0.9256 - val_accuracy: 0.8438\n",
            "Epoch 278/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6819 - accuracy: 0.9698 - val_loss: 0.9364 - val_accuracy: 0.8408\n",
            "Epoch 279/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6726 - accuracy: 0.9717 - val_loss: 0.9069 - val_accuracy: 0.8559\n",
            "Epoch 280/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.6641 - accuracy: 0.9768 - val_loss: 0.8959 - val_accuracy: 0.8619\n",
            "Epoch 281/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6704 - accuracy: 0.9775 - val_loss: 0.9020 - val_accuracy: 0.8664\n",
            "Epoch 282/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6883 - accuracy: 0.9665 - val_loss: 0.9412 - val_accuracy: 0.8438\n",
            "Epoch 283/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6714 - accuracy: 0.9723 - val_loss: 0.8987 - val_accuracy: 0.8544\n",
            "Epoch 284/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6768 - accuracy: 0.9743 - val_loss: 0.9244 - val_accuracy: 0.8453\n",
            "Epoch 285/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6654 - accuracy: 0.9762 - val_loss: 0.9134 - val_accuracy: 0.8604\n",
            "Epoch 286/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6666 - accuracy: 0.9788 - val_loss: 0.8927 - val_accuracy: 0.8634\n",
            "Epoch 287/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6654 - accuracy: 0.9781 - val_loss: 0.9139 - val_accuracy: 0.8483\n",
            "Epoch 288/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6878 - accuracy: 0.9665 - val_loss: 0.9172 - val_accuracy: 0.8559\n",
            "Epoch 289/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6737 - accuracy: 0.9813 - val_loss: 0.9084 - val_accuracy: 0.8423\n",
            "Epoch 290/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6728 - accuracy: 0.9749 - val_loss: 0.9171 - val_accuracy: 0.8468\n",
            "Epoch 291/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6823 - accuracy: 0.9698 - val_loss: 0.8944 - val_accuracy: 0.8604\n",
            "Epoch 292/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6773 - accuracy: 0.9723 - val_loss: 0.9326 - val_accuracy: 0.8423\n",
            "Epoch 293/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6778 - accuracy: 0.9704 - val_loss: 0.8918 - val_accuracy: 0.8544\n",
            "Epoch 294/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6688 - accuracy: 0.9710 - val_loss: 0.9066 - val_accuracy: 0.8438\n",
            "Epoch 295/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6797 - accuracy: 0.9743 - val_loss: 0.9124 - val_accuracy: 0.8498\n",
            "Epoch 296/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6791 - accuracy: 0.9710 - val_loss: 0.9383 - val_accuracy: 0.8529\n",
            "Epoch 297/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6638 - accuracy: 0.9788 - val_loss: 0.9032 - val_accuracy: 0.8559\n",
            "Epoch 298/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6668 - accuracy: 0.9736 - val_loss: 0.9153 - val_accuracy: 0.8574\n",
            "Epoch 299/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6709 - accuracy: 0.9749 - val_loss: 0.8978 - val_accuracy: 0.8634\n",
            "Epoch 300/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6725 - accuracy: 0.9723 - val_loss: 0.8833 - val_accuracy: 0.8679\n",
            "Epoch 301/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6666 - accuracy: 0.9781 - val_loss: 0.8932 - val_accuracy: 0.8544\n",
            "Epoch 302/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6696 - accuracy: 0.9755 - val_loss: 0.8855 - val_accuracy: 0.8619\n",
            "Epoch 303/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6743 - accuracy: 0.9768 - val_loss: 0.9017 - val_accuracy: 0.8619\n",
            "Epoch 304/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6599 - accuracy: 0.9794 - val_loss: 0.8930 - val_accuracy: 0.8664\n",
            "Epoch 305/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6650 - accuracy: 0.9788 - val_loss: 0.8963 - val_accuracy: 0.8649\n",
            "Epoch 306/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6740 - accuracy: 0.9717 - val_loss: 0.9023 - val_accuracy: 0.8574\n",
            "Epoch 307/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6658 - accuracy: 0.9717 - val_loss: 0.9048 - val_accuracy: 0.8498\n",
            "Epoch 308/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6636 - accuracy: 0.9775 - val_loss: 0.9130 - val_accuracy: 0.8498\n",
            "Epoch 309/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6626 - accuracy: 0.9826 - val_loss: 0.9127 - val_accuracy: 0.8468\n",
            "Epoch 310/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6671 - accuracy: 0.9749 - val_loss: 0.9033 - val_accuracy: 0.8408\n",
            "Epoch 311/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6662 - accuracy: 0.9794 - val_loss: 0.9022 - val_accuracy: 0.8559\n",
            "Epoch 312/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6769 - accuracy: 0.9730 - val_loss: 0.9088 - val_accuracy: 0.8664\n",
            "Epoch 313/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6773 - accuracy: 0.9659 - val_loss: 0.9187 - val_accuracy: 0.8483\n",
            "Epoch 314/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6696 - accuracy: 0.9749 - val_loss: 0.9026 - val_accuracy: 0.8679\n",
            "Epoch 315/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6579 - accuracy: 0.9871 - val_loss: 0.9203 - val_accuracy: 0.8498\n",
            "Epoch 316/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6624 - accuracy: 0.9788 - val_loss: 0.8792 - val_accuracy: 0.8634\n",
            "Epoch 317/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6738 - accuracy: 0.9717 - val_loss: 0.8816 - val_accuracy: 0.8634\n",
            "Epoch 318/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6676 - accuracy: 0.9768 - val_loss: 0.8952 - val_accuracy: 0.8498\n",
            "Epoch 319/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6565 - accuracy: 0.9826 - val_loss: 0.8835 - val_accuracy: 0.8769\n",
            "Epoch 320/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6670 - accuracy: 0.9743 - val_loss: 0.8791 - val_accuracy: 0.8724\n",
            "Epoch 321/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6671 - accuracy: 0.9743 - val_loss: 0.8834 - val_accuracy: 0.8589\n",
            "Epoch 322/500\n",
            "49/49 [==============================] - 6s 124ms/step - loss: 0.6615 - accuracy: 0.9762 - val_loss: 0.8744 - val_accuracy: 0.8709\n",
            "Epoch 323/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6649 - accuracy: 0.9755 - val_loss: 0.8898 - val_accuracy: 0.8754\n",
            "Epoch 324/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6646 - accuracy: 0.9736 - val_loss: 0.8934 - val_accuracy: 0.8604\n",
            "Epoch 325/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6606 - accuracy: 0.9736 - val_loss: 0.8876 - val_accuracy: 0.8559\n",
            "Epoch 326/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6674 - accuracy: 0.9710 - val_loss: 0.8931 - val_accuracy: 0.8574\n",
            "Epoch 327/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6781 - accuracy: 0.9659 - val_loss: 0.9176 - val_accuracy: 0.8514\n",
            "Epoch 328/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.6603 - accuracy: 0.9788 - val_loss: 0.8737 - val_accuracy: 0.8589\n",
            "Epoch 329/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6638 - accuracy: 0.9749 - val_loss: 0.8815 - val_accuracy: 0.8604\n",
            "Epoch 330/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6609 - accuracy: 0.9807 - val_loss: 0.8945 - val_accuracy: 0.8649\n",
            "Epoch 331/500\n",
            "49/49 [==============================] - 6s 125ms/step - loss: 0.6513 - accuracy: 0.9813 - val_loss: 0.8644 - val_accuracy: 0.8739\n",
            "Epoch 332/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6632 - accuracy: 0.9749 - val_loss: 0.8893 - val_accuracy: 0.8619\n",
            "Epoch 333/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6564 - accuracy: 0.9788 - val_loss: 0.8749 - val_accuracy: 0.8664\n",
            "Epoch 334/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6632 - accuracy: 0.9801 - val_loss: 0.8982 - val_accuracy: 0.8604\n",
            "Epoch 335/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6624 - accuracy: 0.9820 - val_loss: 0.8922 - val_accuracy: 0.8544\n",
            "Epoch 336/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6620 - accuracy: 0.9781 - val_loss: 0.9013 - val_accuracy: 0.8589\n",
            "Epoch 337/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6638 - accuracy: 0.9794 - val_loss: 0.8892 - val_accuracy: 0.8574\n",
            "Epoch 338/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6569 - accuracy: 0.9788 - val_loss: 0.9083 - val_accuracy: 0.8559\n",
            "Epoch 339/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6618 - accuracy: 0.9807 - val_loss: 0.8919 - val_accuracy: 0.8664\n",
            "Epoch 340/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6562 - accuracy: 0.9755 - val_loss: 0.8716 - val_accuracy: 0.8724\n",
            "Epoch 341/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6601 - accuracy: 0.9743 - val_loss: 0.9011 - val_accuracy: 0.8604\n",
            "Epoch 342/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6571 - accuracy: 0.9775 - val_loss: 0.9106 - val_accuracy: 0.8483\n",
            "Epoch 343/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6683 - accuracy: 0.9743 - val_loss: 0.9074 - val_accuracy: 0.8438\n",
            "Epoch 344/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6572 - accuracy: 0.9788 - val_loss: 0.8901 - val_accuracy: 0.8574\n",
            "Epoch 345/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6465 - accuracy: 0.9884 - val_loss: 0.8746 - val_accuracy: 0.8739\n",
            "Epoch 346/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6494 - accuracy: 0.9871 - val_loss: 0.8883 - val_accuracy: 0.8649\n",
            "Epoch 347/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6639 - accuracy: 0.9788 - val_loss: 0.9046 - val_accuracy: 0.8649\n",
            "Epoch 348/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6643 - accuracy: 0.9717 - val_loss: 0.8723 - val_accuracy: 0.8754\n",
            "Epoch 349/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6532 - accuracy: 0.9788 - val_loss: 0.8791 - val_accuracy: 0.8619\n",
            "Epoch 350/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6618 - accuracy: 0.9723 - val_loss: 0.8987 - val_accuracy: 0.8468\n",
            "Epoch 351/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6506 - accuracy: 0.9826 - val_loss: 0.8917 - val_accuracy: 0.8574\n",
            "Epoch 352/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6527 - accuracy: 0.9788 - val_loss: 0.9342 - val_accuracy: 0.8453\n",
            "Epoch 353/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6439 - accuracy: 0.9839 - val_loss: 0.9064 - val_accuracy: 0.8604\n",
            "Epoch 354/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6567 - accuracy: 0.9807 - val_loss: 0.8754 - val_accuracy: 0.8649\n",
            "Epoch 355/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6496 - accuracy: 0.9826 - val_loss: 0.8868 - val_accuracy: 0.8649\n",
            "Epoch 356/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6621 - accuracy: 0.9775 - val_loss: 0.9081 - val_accuracy: 0.8544\n",
            "Epoch 357/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6684 - accuracy: 0.9749 - val_loss: 0.8977 - val_accuracy: 0.8619\n",
            "Epoch 358/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6632 - accuracy: 0.9762 - val_loss: 0.9155 - val_accuracy: 0.8498\n",
            "Epoch 359/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6425 - accuracy: 0.9826 - val_loss: 0.8891 - val_accuracy: 0.8694\n",
            "Epoch 360/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6524 - accuracy: 0.9801 - val_loss: 0.8784 - val_accuracy: 0.8634\n",
            "Epoch 361/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6430 - accuracy: 0.9839 - val_loss: 0.8749 - val_accuracy: 0.8739\n",
            "Epoch 362/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6655 - accuracy: 0.9723 - val_loss: 0.9104 - val_accuracy: 0.8574\n",
            "Epoch 363/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6458 - accuracy: 0.9807 - val_loss: 0.8724 - val_accuracy: 0.8634\n",
            "Epoch 364/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6509 - accuracy: 0.9833 - val_loss: 0.8956 - val_accuracy: 0.8544\n",
            "Epoch 365/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6422 - accuracy: 0.9839 - val_loss: 0.8886 - val_accuracy: 0.8574\n",
            "Epoch 366/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6461 - accuracy: 0.9788 - val_loss: 0.9014 - val_accuracy: 0.8468\n",
            "Epoch 367/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6513 - accuracy: 0.9794 - val_loss: 0.8852 - val_accuracy: 0.8664\n",
            "Epoch 368/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6633 - accuracy: 0.9743 - val_loss: 0.8994 - val_accuracy: 0.8514\n",
            "Epoch 369/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6652 - accuracy: 0.9768 - val_loss: 0.8650 - val_accuracy: 0.8724\n",
            "Epoch 370/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6591 - accuracy: 0.9762 - val_loss: 0.8979 - val_accuracy: 0.8559\n",
            "Epoch 371/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6424 - accuracy: 0.9871 - val_loss: 0.8878 - val_accuracy: 0.8694\n",
            "Epoch 372/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6597 - accuracy: 0.9775 - val_loss: 0.8910 - val_accuracy: 0.8604\n",
            "Epoch 373/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6396 - accuracy: 0.9852 - val_loss: 0.8752 - val_accuracy: 0.8694\n",
            "Epoch 374/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6430 - accuracy: 0.9801 - val_loss: 0.8767 - val_accuracy: 0.8694\n",
            "Epoch 375/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6564 - accuracy: 0.9743 - val_loss: 0.8691 - val_accuracy: 0.8724\n",
            "Epoch 376/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6561 - accuracy: 0.9807 - val_loss: 0.9047 - val_accuracy: 0.8619\n",
            "Epoch 377/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6471 - accuracy: 0.9852 - val_loss: 0.8857 - val_accuracy: 0.8709\n",
            "Epoch 378/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6513 - accuracy: 0.9794 - val_loss: 0.8921 - val_accuracy: 0.8589\n",
            "Epoch 379/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6520 - accuracy: 0.9801 - val_loss: 0.8738 - val_accuracy: 0.8619\n",
            "Epoch 380/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6522 - accuracy: 0.9781 - val_loss: 0.8910 - val_accuracy: 0.8544\n",
            "Epoch 381/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6344 - accuracy: 0.9865 - val_loss: 0.8901 - val_accuracy: 0.8604\n",
            "Epoch 382/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6439 - accuracy: 0.9807 - val_loss: 0.9029 - val_accuracy: 0.8363\n",
            "Epoch 383/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6450 - accuracy: 0.9813 - val_loss: 0.8824 - val_accuracy: 0.8814\n",
            "Epoch 384/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6387 - accuracy: 0.9884 - val_loss: 0.8737 - val_accuracy: 0.8724\n",
            "Epoch 385/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6491 - accuracy: 0.9801 - val_loss: 0.8931 - val_accuracy: 0.8559\n",
            "Epoch 386/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6488 - accuracy: 0.9807 - val_loss: 0.8666 - val_accuracy: 0.8634\n",
            "Epoch 387/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6424 - accuracy: 0.9846 - val_loss: 0.8880 - val_accuracy: 0.8634\n",
            "Epoch 388/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6341 - accuracy: 0.9839 - val_loss: 0.8939 - val_accuracy: 0.8544\n",
            "Epoch 389/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6471 - accuracy: 0.9781 - val_loss: 0.8944 - val_accuracy: 0.8619\n",
            "Epoch 390/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6463 - accuracy: 0.9755 - val_loss: 0.8938 - val_accuracy: 0.8694\n",
            "Epoch 391/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6453 - accuracy: 0.9775 - val_loss: 0.8764 - val_accuracy: 0.8754\n",
            "Epoch 392/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6439 - accuracy: 0.9839 - val_loss: 0.8998 - val_accuracy: 0.8604\n",
            "Epoch 393/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6511 - accuracy: 0.9813 - val_loss: 0.8829 - val_accuracy: 0.8544\n",
            "Epoch 394/500\n",
            "49/49 [==============================] - 6s 130ms/step - loss: 0.6505 - accuracy: 0.9813 - val_loss: 0.8614 - val_accuracy: 0.8769\n",
            "Epoch 395/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6611 - accuracy: 0.9755 - val_loss: 0.8789 - val_accuracy: 0.8679\n",
            "Epoch 396/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6536 - accuracy: 0.9846 - val_loss: 0.8787 - val_accuracy: 0.8679\n",
            "Epoch 397/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6376 - accuracy: 0.9813 - val_loss: 0.8989 - val_accuracy: 0.8619\n",
            "Epoch 398/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6528 - accuracy: 0.9801 - val_loss: 0.8636 - val_accuracy: 0.8784\n",
            "Epoch 399/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6452 - accuracy: 0.9820 - val_loss: 0.8696 - val_accuracy: 0.8649\n",
            "Epoch 400/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6386 - accuracy: 0.9852 - val_loss: 0.9021 - val_accuracy: 0.8514\n",
            "Epoch 401/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6432 - accuracy: 0.9788 - val_loss: 0.8923 - val_accuracy: 0.8574\n",
            "Epoch 402/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6461 - accuracy: 0.9839 - val_loss: 0.8747 - val_accuracy: 0.8754\n",
            "Epoch 403/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6482 - accuracy: 0.9807 - val_loss: 0.8750 - val_accuracy: 0.8769\n",
            "Epoch 404/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6372 - accuracy: 0.9871 - val_loss: 0.8727 - val_accuracy: 0.8649\n",
            "Epoch 405/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6504 - accuracy: 0.9794 - val_loss: 0.8704 - val_accuracy: 0.8679\n",
            "Epoch 406/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6386 - accuracy: 0.9852 - val_loss: 0.8918 - val_accuracy: 0.8574\n",
            "Epoch 407/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6523 - accuracy: 0.9813 - val_loss: 0.8960 - val_accuracy: 0.8619\n",
            "Epoch 408/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6492 - accuracy: 0.9807 - val_loss: 0.8489 - val_accuracy: 0.8814\n",
            "Epoch 409/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6461 - accuracy: 0.9826 - val_loss: 0.8490 - val_accuracy: 0.8874\n",
            "Epoch 410/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6423 - accuracy: 0.9897 - val_loss: 0.8707 - val_accuracy: 0.8679\n",
            "Epoch 411/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6456 - accuracy: 0.9813 - val_loss: 0.8815 - val_accuracy: 0.8649\n",
            "Epoch 412/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6422 - accuracy: 0.9820 - val_loss: 0.9033 - val_accuracy: 0.8604\n",
            "Epoch 413/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6397 - accuracy: 0.9839 - val_loss: 0.8693 - val_accuracy: 0.8694\n",
            "Epoch 414/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6391 - accuracy: 0.9826 - val_loss: 0.8871 - val_accuracy: 0.8754\n",
            "Epoch 415/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6438 - accuracy: 0.9833 - val_loss: 0.8889 - val_accuracy: 0.8589\n",
            "Epoch 416/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6479 - accuracy: 0.9820 - val_loss: 0.8913 - val_accuracy: 0.8694\n",
            "Epoch 417/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6472 - accuracy: 0.9826 - val_loss: 0.9014 - val_accuracy: 0.8423\n",
            "Epoch 418/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6444 - accuracy: 0.9801 - val_loss: 0.8768 - val_accuracy: 0.8724\n",
            "Epoch 419/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6378 - accuracy: 0.9852 - val_loss: 0.8832 - val_accuracy: 0.8574\n",
            "Epoch 420/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6388 - accuracy: 0.9871 - val_loss: 0.8730 - val_accuracy: 0.8799\n",
            "Epoch 421/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6352 - accuracy: 0.9865 - val_loss: 0.8577 - val_accuracy: 0.8859\n",
            "Epoch 422/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6389 - accuracy: 0.9839 - val_loss: 0.8965 - val_accuracy: 0.8634\n",
            "Epoch 423/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6318 - accuracy: 0.9910 - val_loss: 0.8876 - val_accuracy: 0.8619\n",
            "Epoch 424/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6370 - accuracy: 0.9865 - val_loss: 0.8840 - val_accuracy: 0.8529\n",
            "Epoch 425/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6357 - accuracy: 0.9833 - val_loss: 0.8972 - val_accuracy: 0.8649\n",
            "Epoch 426/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6454 - accuracy: 0.9820 - val_loss: 0.8701 - val_accuracy: 0.8769\n",
            "Epoch 427/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6391 - accuracy: 0.9839 - val_loss: 0.8802 - val_accuracy: 0.8589\n",
            "Epoch 428/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6357 - accuracy: 0.9858 - val_loss: 0.8760 - val_accuracy: 0.8589\n",
            "Epoch 429/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6384 - accuracy: 0.9794 - val_loss: 0.8825 - val_accuracy: 0.8559\n",
            "Epoch 430/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6434 - accuracy: 0.9807 - val_loss: 0.8844 - val_accuracy: 0.8619\n",
            "Epoch 431/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6518 - accuracy: 0.9801 - val_loss: 0.8796 - val_accuracy: 0.8694\n",
            "Epoch 432/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6387 - accuracy: 0.9891 - val_loss: 0.8711 - val_accuracy: 0.8679\n",
            "Epoch 433/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6519 - accuracy: 0.9788 - val_loss: 0.9073 - val_accuracy: 0.8498\n",
            "Epoch 434/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6470 - accuracy: 0.9807 - val_loss: 0.8942 - val_accuracy: 0.8694\n",
            "Epoch 435/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6332 - accuracy: 0.9846 - val_loss: 0.8739 - val_accuracy: 0.8709\n",
            "Epoch 436/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6437 - accuracy: 0.9826 - val_loss: 0.8649 - val_accuracy: 0.8784\n",
            "Epoch 437/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6391 - accuracy: 0.9833 - val_loss: 0.9089 - val_accuracy: 0.8453\n",
            "Epoch 438/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6366 - accuracy: 0.9871 - val_loss: 0.8736 - val_accuracy: 0.8619\n",
            "Epoch 439/500\n",
            "49/49 [==============================] - 6s 116ms/step - loss: 0.6394 - accuracy: 0.9858 - val_loss: 0.8661 - val_accuracy: 0.8754\n",
            "Epoch 440/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6348 - accuracy: 0.9865 - val_loss: 0.8799 - val_accuracy: 0.8679\n",
            "Epoch 441/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6380 - accuracy: 0.9807 - val_loss: 0.8774 - val_accuracy: 0.8784\n",
            "Epoch 442/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6342 - accuracy: 0.9858 - val_loss: 0.8776 - val_accuracy: 0.8649\n",
            "Epoch 443/500\n",
            "49/49 [==============================] - 6s 126ms/step - loss: 0.6288 - accuracy: 0.9884 - val_loss: 0.8820 - val_accuracy: 0.8664\n",
            "Epoch 444/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6251 - accuracy: 0.9891 - val_loss: 0.9061 - val_accuracy: 0.8589\n",
            "Epoch 445/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6458 - accuracy: 0.9801 - val_loss: 0.8631 - val_accuracy: 0.8754\n",
            "Epoch 446/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6331 - accuracy: 0.9852 - val_loss: 0.8749 - val_accuracy: 0.8619\n",
            "Epoch 447/500\n",
            "49/49 [==============================] - 6s 117ms/step - loss: 0.6303 - accuracy: 0.9891 - val_loss: 0.8592 - val_accuracy: 0.8769\n",
            "Epoch 448/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6371 - accuracy: 0.9891 - val_loss: 0.8841 - val_accuracy: 0.8679\n",
            "Epoch 449/500\n",
            "49/49 [==============================] - 6s 118ms/step - loss: 0.6434 - accuracy: 0.9813 - val_loss: 0.8819 - val_accuracy: 0.8739\n",
            "Epoch 450/500\n",
            "49/49 [==============================] - 6s 119ms/step - loss: 0.6337 - accuracy: 0.9852 - val_loss: 0.8852 - val_accuracy: 0.8544\n",
            "Epoch 451/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6333 - accuracy: 0.9878 - val_loss: 0.8684 - val_accuracy: 0.8694\n",
            "Epoch 452/500\n",
            "49/49 [==============================] - 6s 124ms/step - loss: 0.6248 - accuracy: 0.9910 - val_loss: 0.8946 - val_accuracy: 0.8589\n",
            "Epoch 453/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6327 - accuracy: 0.9897 - val_loss: 0.8629 - val_accuracy: 0.8754\n",
            "Epoch 454/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6285 - accuracy: 0.9878 - val_loss: 0.8674 - val_accuracy: 0.8754\n",
            "Epoch 455/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6369 - accuracy: 0.9878 - val_loss: 0.8841 - val_accuracy: 0.8769\n",
            "Epoch 456/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6270 - accuracy: 0.9884 - val_loss: 0.8742 - val_accuracy: 0.8679\n",
            "Epoch 457/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6282 - accuracy: 0.9897 - val_loss: 0.8819 - val_accuracy: 0.8649\n",
            "Epoch 458/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6240 - accuracy: 0.9891 - val_loss: 0.8763 - val_accuracy: 0.8664\n",
            "Epoch 459/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6377 - accuracy: 0.9826 - val_loss: 0.8602 - val_accuracy: 0.8709\n",
            "Epoch 460/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6301 - accuracy: 0.9858 - val_loss: 0.8814 - val_accuracy: 0.8664\n",
            "Epoch 461/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6324 - accuracy: 0.9897 - val_loss: 0.8776 - val_accuracy: 0.8694\n",
            "Epoch 462/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6333 - accuracy: 0.9871 - val_loss: 0.8812 - val_accuracy: 0.8604\n",
            "Epoch 463/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6279 - accuracy: 0.9891 - val_loss: 0.8755 - val_accuracy: 0.8829\n",
            "Epoch 464/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6304 - accuracy: 0.9884 - val_loss: 0.8774 - val_accuracy: 0.8829\n",
            "Epoch 465/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6418 - accuracy: 0.9839 - val_loss: 0.8744 - val_accuracy: 0.8739\n",
            "Epoch 466/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6356 - accuracy: 0.9897 - val_loss: 0.8890 - val_accuracy: 0.8679\n",
            "Epoch 467/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6419 - accuracy: 0.9775 - val_loss: 0.8752 - val_accuracy: 0.8784\n",
            "Epoch 468/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6241 - accuracy: 0.9858 - val_loss: 0.8888 - val_accuracy: 0.8559\n",
            "Epoch 469/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6369 - accuracy: 0.9858 - val_loss: 0.8690 - val_accuracy: 0.8649\n",
            "Epoch 470/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6266 - accuracy: 0.9916 - val_loss: 0.8909 - val_accuracy: 0.8634\n",
            "Epoch 471/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6284 - accuracy: 0.9871 - val_loss: 0.9028 - val_accuracy: 0.8453\n",
            "Epoch 472/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6243 - accuracy: 0.9891 - val_loss: 0.8818 - val_accuracy: 0.8619\n",
            "Epoch 473/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6289 - accuracy: 0.9884 - val_loss: 0.8568 - val_accuracy: 0.8844\n",
            "Epoch 474/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6292 - accuracy: 0.9891 - val_loss: 0.8861 - val_accuracy: 0.8604\n",
            "Epoch 475/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6279 - accuracy: 0.9878 - val_loss: 0.9012 - val_accuracy: 0.8498\n",
            "Epoch 476/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6384 - accuracy: 0.9820 - val_loss: 0.8693 - val_accuracy: 0.8724\n",
            "Epoch 477/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6248 - accuracy: 0.9923 - val_loss: 0.8942 - val_accuracy: 0.8589\n",
            "Epoch 478/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6395 - accuracy: 0.9788 - val_loss: 0.8868 - val_accuracy: 0.8483\n",
            "Epoch 479/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6324 - accuracy: 0.9833 - val_loss: 0.8872 - val_accuracy: 0.8589\n",
            "Epoch 480/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6340 - accuracy: 0.9858 - val_loss: 0.8832 - val_accuracy: 0.8649\n",
            "Epoch 481/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6390 - accuracy: 0.9826 - val_loss: 0.8673 - val_accuracy: 0.8784\n",
            "Epoch 482/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6297 - accuracy: 0.9871 - val_loss: 0.8992 - val_accuracy: 0.8408\n",
            "Epoch 483/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6349 - accuracy: 0.9878 - val_loss: 0.8909 - val_accuracy: 0.8498\n",
            "Epoch 484/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6341 - accuracy: 0.9865 - val_loss: 0.8922 - val_accuracy: 0.8574\n",
            "Epoch 485/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6407 - accuracy: 0.9801 - val_loss: 0.8839 - val_accuracy: 0.8589\n",
            "Epoch 486/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6233 - accuracy: 0.9865 - val_loss: 0.8868 - val_accuracy: 0.8619\n",
            "Epoch 487/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6232 - accuracy: 0.9949 - val_loss: 0.8861 - val_accuracy: 0.8604\n",
            "Epoch 488/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6278 - accuracy: 0.9903 - val_loss: 0.8832 - val_accuracy: 0.8574\n",
            "Epoch 489/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6278 - accuracy: 0.9910 - val_loss: 0.8775 - val_accuracy: 0.8634\n",
            "Epoch 490/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6282 - accuracy: 0.9884 - val_loss: 0.8831 - val_accuracy: 0.8483\n",
            "Epoch 491/500\n",
            "49/49 [==============================] - 6s 124ms/step - loss: 0.6207 - accuracy: 0.9897 - val_loss: 0.8931 - val_accuracy: 0.8574\n",
            "Epoch 492/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6291 - accuracy: 0.9884 - val_loss: 0.8694 - val_accuracy: 0.8709\n",
            "Epoch 493/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6235 - accuracy: 0.9923 - val_loss: 0.8891 - val_accuracy: 0.8634\n",
            "Epoch 494/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6211 - accuracy: 0.9916 - val_loss: 0.8857 - val_accuracy: 0.8574\n",
            "Epoch 495/500\n",
            "49/49 [==============================] - 6s 122ms/step - loss: 0.6184 - accuracy: 0.9923 - val_loss: 0.8682 - val_accuracy: 0.8844\n",
            "Epoch 496/500\n",
            "49/49 [==============================] - 6s 123ms/step - loss: 0.6314 - accuracy: 0.9839 - val_loss: 0.9177 - val_accuracy: 0.8514\n",
            "Epoch 497/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6320 - accuracy: 0.9865 - val_loss: 0.8965 - val_accuracy: 0.8574\n",
            "Epoch 498/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6242 - accuracy: 0.9878 - val_loss: 0.8866 - val_accuracy: 0.8649\n",
            "Epoch 499/500\n",
            "49/49 [==============================] - 6s 120ms/step - loss: 0.6196 - accuracy: 0.9903 - val_loss: 0.8683 - val_accuracy: 0.8649\n",
            "Epoch 500/500\n",
            "49/49 [==============================] - 6s 121ms/step - loss: 0.6312 - accuracy: 0.9846 - val_loss: 0.8619 - val_accuracy: 0.8769\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "SOiKl3mrapF9",
        "outputId": "fa7b2bce-562a-43f0-b10c-acf6b328d61c"
      },
      "source": [
        "plt.plot(r.history['loss'], label='train loss')\n",
        "plt.plot(r.history['val_loss'], label='val loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 112,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "us_8rj48gwPZ"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}