0528_LeeJaeho.ipynb

{
  "metadata": {
    "kernelspec": {
      "language": "python",
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "version": "3.6.4",
      "file_extension": ".py",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "name": "python",
      "mimetype": "text/x-python"
    },
    "colab": {
      "provenance": [],
      "gpuType": "T4"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat_minor": 0,
  "nbformat": 4,
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "import os\n",
        "import cv2\n",
        "import numpy as np \n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from pathlib import Path\n",
        "from collections import Counter\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers\n",
        "print(\"Tensorflow version: \", tf.__version__)\n",
        "\n",
        "seed = 1234\n",
        "np.random.seed(seed)\n",
        "tf.random.set_seed(seed)"
      ],
      "metadata": {
        "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
        "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MsmJtNa34A42",
        "outputId": "0536798f-ecd6-4805-931f-6915eb8a7c98"
      },
      "execution_count": 54,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Tensorflow version:  2.12.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!dir"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YFQTi1GLOJjw",
        "outputId": "678ae774-902c-42cf-a1fb-e850df33ed2c"
      },
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "drive  sample_data\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4yGS0iRROPkP",
        "outputId": "3f5df92d-a124-4259-d1b2-c321559806a4"
      },
      "execution_count": 56,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Path to the data directory\n",
        "\n",
        "# train 모델의 캡챠 글자 개수 결정.\n",
        "captcha=5\n",
        "\n",
        "# images : 글자 4개\n",
        "# samples : 글자 5개\n",
        "data_dir = Path(\"/content/drive/MyDrive/input/archive\")\n",
        "\n",
        "# Get list of all the images\n",
        "images = list(data_dir.glob(\"*.png\"))\n",
        "images += list(data_dir.glob(\"*.jpg\"))\n",
        "print(\"Number of images found: \", len(images))\n",
        "\n",
        "\n",
        "# Let's take a look at some samples first. \n",
        "# Always look at your data!\n",
        "sample_images = images[:4]\n",
        "\n",
        "_,ax = plt.subplots(2,2, figsize=(5,3))\n",
        "for i in range(4):\n",
        "    img = cv2.imread(str(sample_images[i]))\n",
        "    print(\"Shape of image: \", img.shape)\n",
        "    ax[i//2, i%2].imshow(img)\n",
        "    ax[i//2, i%2].axis('off')\n",
        "plt.show()"
      ],
      "metadata": {
        "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a",
        "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "1bIFmJPl4A45",
        "outputId": "90777edc-0708-4830-e1e6-8a02039de6ea"
      },
      "execution_count": 84,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of images found:  1070\n",
            "Shape of image:  (40, 150, 3)\n",
            "Shape of image:  (40, 150, 3)\n",
            "Shape of image:  (40, 150, 3)\n",
            "Shape of image:  (40, 150, 3)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x300 with 4 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Store all the characters in a set\n",
        "characters = set()\n",
        "\n",
        "# A list to store the length of each captcha\n",
        "captcha_length = []\n",
        "\n",
        "# Store image-label info\n",
        "dataset = []\n",
        "\n",
        "# Iterate over the dataset and store the\n",
        "# information needed\n",
        "for img_path in images:\n",
        "\n",
        "  if str(img_path)[-3:] == 'jpg':\n",
        "    label = img_path.name.split(\".jpg\")[0]\n",
        "  else:\n",
        "    label = img_path.name.split(\".png\")[0]\n",
        "  # 2. Store the length of this cpatcha\n",
        "  captcha_length.append(len(label))\n",
        "  # 3. Store the image-label pair info\n",
        "  dataset.append((str(img_path), label))\n",
        "    \n",
        "  # 4. Store the characters present\n",
        "  for ch in label:\n",
        "    characters.add(ch)\n",
        "\n",
        "# Sort the characters        \n",
        "characters = sorted(characters)\n",
        "\n",
        "# Convert the dataset info into a dataframe\n",
        "dataset = pd.DataFrame(dataset, columns=[\"img_path\", \"label\"], index=None)\n",
        "\n",
        "# Shuffle the dataset\n",
        "dataset = dataset.sample(frac=1.).reset_index(drop=True)\n",
        "\n",
        "\n",
        "print(\"Number of unqiue charcaters in the whole dataset: \", len(characters))\n",
        "print(\"Maximum length of any captcha: \", max(Counter(captcha_length).keys()))\n",
        "print(\"Characters present: \", characters)\n",
        "print(\"Total number of samples in the dataset: \", len(dataset))\n",
        "dataset.head()"
      ],
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 299
        },
        "id": "sfacnY4d4A46",
        "outputId": "601f7458-9982-4bd0-ff59-df3c5de48384"
      },
      "execution_count": 85,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of unqiue charcaters in the whole dataset:  60\n",
            "Maximum length of any captcha:  5\n",
            "Characters present:  ['1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']\n",
            "Total number of samples in the dataset:  1070\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                         img_path  label\n",
              "0  /content/drive/MyDrive/input/archive/zsCiK.jpg  zsCiK\n",
              "1  /content/drive/MyDrive/input/archive/zWM8S.jpg  zWM8S\n",
              "2  /content/drive/MyDrive/input/archive/zZb3T.jpg  zZb3T\n",
              "3  /content/drive/MyDrive/input/archive/zTP3a.jpg  zTP3a\n",
              "4  /content/drive/MyDrive/input/archive/zaUbf.jpg  zaUbf"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-380c80e1-bdeb-4785-80f0-105eb3221563\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>img_path</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>/content/drive/MyDrive/input/archive/zsCiK.jpg</td>\n",
              "      <td>zsCiK</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>/content/drive/MyDrive/input/archive/zWM8S.jpg</td>\n",
              "      <td>zWM8S</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>/content/drive/MyDrive/input/archive/zZb3T.jpg</td>\n",
              "      <td>zZb3T</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>/content/drive/MyDrive/input/archive/zTP3a.jpg</td>\n",
              "      <td>zTP3a</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>/content/drive/MyDrive/input/archive/zaUbf.jpg</td>\n",
              "      <td>zaUbf</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-380c80e1-bdeb-4785-80f0-105eb3221563')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-380c80e1-bdeb-4785-80f0-105eb3221563 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-380c80e1-bdeb-4785-80f0-105eb3221563');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 85
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Split the dataset into training and validation sets\n",
        "training_data, validation_data = train_test_split(dataset, test_size=0.1, random_state=seed)\n",
        "\n",
        "training_data = training_data.reset_index(drop=True)\n",
        "validation_data = validation_data.reset_index(drop=True)\n",
        "\n",
        "print(\"Number of training samples: \", len(training_data))\n",
        "print(\"Number of validation samples: \", len(validation_data))\n",
        "\n",
        "\n",
        "\n",
        "# Map text to numeric labels \n",
        "char_to_labels = {char:idx for idx, char in enumerate(characters)}\n",
        "\n",
        "# Map numeric labels to text\n",
        "labels_to_char = {val:key for key, val in char_to_labels.items()}\n",
        "\n",
        "\n",
        "\n",
        "# Sanity check for corrupted images\n",
        "def is_valid_captcha(captcha):\n",
        "    for ch in captcha:\n",
        "        if not ch in characters:\n",
        "            return False\n",
        "    return True\n",
        "\n",
        "\n",
        "\n",
        "# Store arrays in memory as it's not a much big dataset\n",
        "def generate_arrays(df, resize=True, img_height=50, img_width=200):\n",
        "    \"\"\"Generates image array and labels array from a dataframe.\n",
        "    \n",
        "    Args:\n",
        "        df: dataframe from which we want to read the data\n",
        "        resize (bool)    : whether to resize images or not\n",
        "        img_weidth (int): width of the resized images\n",
        "        img_height (int): height of the resized images\n",
        "        \n",
        "    Returns:\n",
        "        images (ndarray): grayscale images\n",
        "        labels (ndarray): corresponding encoded labels\n",
        "    \"\"\"\n",
        "    \n",
        "    num_items = len(df)\n",
        "    images = np.zeros((num_items, img_height, img_width), dtype=np.float32)\n",
        "    labels = [0]*num_items\n",
        "    \n",
        "    for i in range(num_items):\n",
        "        img = cv2.imread(df[\"img_path\"][i])\n",
        "        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n",
        "        \n",
        "        if resize: \n",
        "            img = cv2.resize(img, (img_width, img_height))\n",
        "        \n",
        "        img = (img/255.).astype(np.float32)\n",
        "        label = df[\"label\"][i]\n",
        "        \n",
        "        # Add only if it is a valid captcha\n",
        "        if is_valid_captcha(label):\n",
        "            images[i, :, :] = img\n",
        "            labels[i] = label\n",
        "    \n",
        "    return images, np.array(labels)\n",
        "\n",
        "\n",
        "\n",
        "# Build training data\n",
        "training_data, training_labels = generate_arrays(df=training_data)\n",
        "print(\"Number of training images: \", training_data.shape)\n",
        "print(\"Number of training labels: \", training_labels.shape)\n",
        "\n",
        "\n",
        "# Build validation data\n",
        "validation_data, validation_labels = generate_arrays(df=validation_data)\n",
        "print(\"Number of validation images: \", validation_data.shape)\n",
        "print(\"Number of validation labels: \", validation_labels.shape)"
      ],
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ecL0v4hx4A47",
        "outputId": "0a2a0f6f-c86a-43f2-c788-ecfe683620fc"
      },
      "execution_count": 86,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of training samples:  963\n",
            "Number of validation samples:  107\n",
            "Number of training images:  (963, 50, 200)\n",
            "Number of training labels:  (963,)\n",
            "Number of validation images:  (107, 50, 200)\n",
            "Number of validation labels:  (107,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class DataGenerator(keras.utils.Sequence):\n",
        "    \"\"\"Generates batches from a given dataset.\n",
        "    \n",
        "    Args:\n",
        "        data: training or validation data\n",
        "        labels: corresponding labels\n",
        "        char_map: dictionary mapping char to labels\n",
        "        batch_size: size of a single batch\n",
        "        img_width: width of the resized\n",
        "        img_height: height of the resized\n",
        "        downsample_factor: by what factor did the CNN downsample the images\n",
        "        max_length: maximum length of any captcha\n",
        "        shuffle: whether to shuffle data or not after each epoch\n",
        "    Returns:\n",
        "        batch_inputs: a dictionary containing batch inputs \n",
        "        batch_labels: a batch of corresponding labels \n",
        "    \"\"\"\n",
        "    \n",
        "    def __init__(self,\n",
        "                 data,\n",
        "                 labels,\n",
        "                 char_map,\n",
        "                 batch_size=16,\n",
        "                 img_width=200,\n",
        "                 img_height=50,\n",
        "                 downsample_factor=4,\n",
        "                 max_length=5, # 글자 개수에 맞춰서 4개 or 5개\n",
        "                 shuffle=True\n",
        "                ):\n",
        "        self.data = data\n",
        "        self.labels = labels\n",
        "        self.char_map = char_map\n",
        "        self.batch_size = batch_size\n",
        "        self.img_width = img_width\n",
        "        self.img_height = img_height\n",
        "        self.downsample_factor = downsample_factor\n",
        "        self.max_length = max_length\n",
        "        self.shuffle = shuffle\n",
        "        self.indices = np.arange(len(data))    \n",
        "        self.on_epoch_end()\n",
        "        \n",
        "    def __len__(self):\n",
        "        return int(np.ceil(len(self.data) / self.batch_size))\n",
        "    \n",
        "    def __getitem__(self, idx):\n",
        "        # 1. Get the next batch indices\n",
        "        curr_batch_idx = self.indices[idx*self.batch_size:(idx+1)*self.batch_size]\n",
        "        \n",
        "        # 2. This isn't necessary but it can help us save some memory\n",
        "        # as not all batches the last batch may not have elements\n",
        "        # equal to the batch_size \n",
        "        batch_len = len(curr_batch_idx)\n",
        "        \n",
        "        # 3. Instantiate batch arrays\n",
        "        batch_images = np.ones((batch_len, self.img_width, self.img_height, 1),\n",
        "                               dtype=np.float32)\n",
        "        batch_labels = np.ones((batch_len, self.max_length), dtype=np.float32)\n",
        "        input_length = np.ones((batch_len, 1), dtype=np.int64) * \\\n",
        "                                (self.img_width // self.downsample_factor - 2)\n",
        "        label_length = np.zeros((batch_len, 1), dtype=np.int64)\n",
        "        \n",
        "        \n",
        "        for j, idx in enumerate(curr_batch_idx):\n",
        "            # 1. Get the image and transpose it\n",
        "            img = self.data[idx].T\n",
        "            # 2. Add extra dimenison\n",
        "            img = np.expand_dims(img, axis=-1)\n",
        "            # 3. Get the correpsonding label\n",
        "            text = self.labels[idx]\n",
        "            # 4. Include the pair only if the captcha is valid\n",
        "            if is_valid_captcha(text):\n",
        "                label = [self.char_map[ch] for ch in text]\n",
        "                batch_images[j] = img\n",
        "                batch_labels[j] = label\n",
        "                label_length[j] = len(text)\n",
        "        \n",
        "        batch_inputs = {\n",
        "                'input_data': batch_images,\n",
        "                'input_label': batch_labels,\n",
        "                'input_length': input_length,\n",
        "                'label_length': label_length,\n",
        "                }\n",
        "        return batch_inputs, np.zeros(batch_len).astype(np.float32)\n",
        "        \n",
        "    \n",
        "    def on_epoch_end(self):\n",
        "        if self.shuffle:\n",
        "            np.random.shuffle(self.indices)"
      ],
      "metadata": {
        "trusted": true,
        "id": "yOQu8z0r4A48"
      },
      "execution_count": 87,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Batch size for training and validation\n",
        "batch_size = 32\n",
        "\n",
        "# Desired image dimensions\n",
        "img_width=200\n",
        "img_height=50 \n",
        "\n",
        "# Factor  by which the image is going to be downsampled\n",
        "# by the convolutional blocks\n",
        "downsample_factor=4\n",
        "\n",
        "# Maximum length of any captcha in the data\n",
        "# 글자 개수에 맞춰서 4개 or 5개\n",
        "max_length = captcha\n",
        "\n",
        "# Get a generator object for the training data\n",
        "train_data_generator = DataGenerator(data=training_data,\n",
        "                                     labels=training_labels,\n",
        "                                     char_map=char_to_labels,\n",
        "                                     batch_size=batch_size,\n",
        "                                     img_width=img_width,\n",
        "                                     img_height=img_height,\n",
        "                                     downsample_factor=downsample_factor,\n",
        "                                     max_length=max_length,\n",
        "                                     shuffle=True\n",
        "                                    )\n",
        "\n",
        "# Get a generator object for the validation data \n",
        "valid_data_generator = DataGenerator(data=validation_data,\n",
        "                                     labels=validation_labels,\n",
        "                                     char_map=char_to_labels,\n",
        "                                     batch_size=batch_size,\n",
        "                                     img_width=img_width,\n",
        "                                     img_height=img_height,\n",
        "                                     downsample_factor=downsample_factor,\n",
        "                                     max_length=max_length,\n",
        "                                     shuffle=False\n",
        "                                    )"
      ],
      "metadata": {
        "trusted": true,
        "id": "FjlSMQX24A49"
      },
      "execution_count": 88,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class CTCLayer(layers.Layer):\n",
        "    def __init__(self, name=None):\n",
        "        super().__init__(name=name)\n",
        "        self.loss_fn = keras.backend.ctc_batch_cost\n",
        "\n",
        "    def call(self, y_true, y_pred, input_length, label_length):\n",
        "        # Compute the training-time loss value and add it\n",
        "        # to the layer using `self.add_loss()`.\n",
        "        loss = self.loss_fn(y_true, y_pred, input_length, label_length)\n",
        "        self.add_loss(loss)\n",
        "        \n",
        "        # On test time, just return the computed loss\n",
        "        return loss\n",
        "\n",
        "\n",
        "\n",
        "def build_model(opt, lr=0.001):\n",
        "    # Inputs to the model\n",
        "    input_img = layers.Input(shape=(img_width, img_height, 1),\n",
        "                            name='input_data',\n",
        "                            dtype='float32')\n",
        "    labels = layers.Input(name='input_label', shape=[max_length], dtype='float32')\n",
        "    input_length = layers.Input(name='input_length', shape=[1], dtype='int64')\n",
        "    label_length = layers.Input(name='label_length', shape=[1], dtype='int64')\n",
        "    \n",
        "    # First conv block\n",
        "    x = layers.Conv2D(32,\n",
        "               (3,3),\n",
        "               activation='relu',\n",
        "               kernel_initializer='he_normal',\n",
        "               padding='same',\n",
        "               name='Conv1')(input_img)\n",
        "    x = layers.MaxPooling2D((2,2), name='pool1')(x)\n",
        "    \n",
        "    # Second conv block\n",
        "    x = layers.Conv2D(64,\n",
        "               (3,3),\n",
        "               activation='relu',\n",
        "               kernel_initializer='he_normal',\n",
        "               padding='same',\n",
        "               name='Conv2')(x)\n",
        "    x = layers.MaxPooling2D((2,2), name='pool2')(x)\n",
        "    \n",
        "    # We have used two max pool with pool size and strides of 2.\n",
        "    # Hence, downsampled feature maps are 4x smaller. The number of\n",
        "    # filters in the last layer is 64. Reshape accordingly before\n",
        "    # passing it to RNNs\n",
        "    new_shape = ((img_width // 4), (img_height // 4)*64)\n",
        "    x = layers.Reshape(target_shape=new_shape, name='reshape')(x)\n",
        "    x = layers.Dense(64, activation='relu', name='dense1')(x)\n",
        "    x = layers.Dropout(0.2)(x)\n",
        "    \n",
        "    # RNNs\n",
        "    x = layers.Bidirectional(layers.LSTM(128,\n",
        "                                         return_sequences=True,\n",
        "                                         dropout=0.2))(x)\n",
        "    x = layers.Bidirectional(layers.LSTM(64,\n",
        "                                         return_sequences=True,\n",
        "                                         dropout=0.25))(x)\n",
        "    \n",
        "    # Predictions\n",
        "    x = layers.Dense(len(characters)+1,\n",
        "              activation='softmax', \n",
        "              name='dense2',\n",
        "              kernel_initializer='he_normal')(x)\n",
        "    \n",
        "    # Calculate CTC\n",
        "    output = CTCLayer(name='ctc_loss')(labels, x, input_length, label_length)\n",
        "    \n",
        "    # Define the model\n",
        "    model = keras.models.Model(inputs=[input_img,\n",
        "                                       labels,\n",
        "                                       input_length,\n",
        "                                       label_length],\n",
        "                                outputs=output,\n",
        "                                name='ocr_model_v1')\n",
        "    \n",
        "    # Optimizer\n",
        "    sgd = keras.optimizers.legacy.SGD(learning_rate=lr,\n",
        "                               decay=1e-6,\n",
        "                               momentum=0.9,\n",
        "                               nesterov=True,\n",
        "                               clipnorm=5)\n",
        "    \n",
        "    Adam = tf.keras.optimizers.Adam(\n",
        "        learning_rate=lr,\n",
        "        beta_1=0.9,\n",
        "        beta_2=0.999,\n",
        "        epsilon=1e-07,\n",
        "        amsgrad=False,\n",
        "        weight_decay=None,\n",
        "        clipnorm=None,\n",
        "        clipvalue=None,\n",
        "        global_clipnorm=None,\n",
        "        use_ema=False,\n",
        "        ema_momentum=0.99,\n",
        "        ema_overwrite_frequency=None,\n",
        "        jit_compile=True)\n",
        "\n",
        "    if opt=='SGD':\n",
        "      opt=sgd\n",
        "\n",
        "    if opt=='Adam':\n",
        "      opt=Adam  \n",
        "      \n",
        "    # Compile the model and return\n",
        "    model.compile(optimizer=opt, metrics=['accuracy'])\n",
        "    return model"
      ],
      "metadata": {
        "trusted": true,
        "id": "Xo6-sWZI4A4-"
      },
      "execution_count": 89,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = build_model(opt='Adam', lr=0.001)\n",
        "# Add early stopping\n",
        "es = keras.callbacks.EarlyStopping(monitor='val_loss',\n",
        "                                        patience=5,\n",
        "                                        restore_best_weights=True)\n",
        "\n",
        "# Train the model\n",
        "history = model.fit_generator(train_data_generator,\n",
        "                          validation_data=valid_data_generator,\n",
        "                          epochs=100,\n",
        "                          callbacks=[es])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eqKxzNQ6EiJI",
        "outputId": "53636795-94aa-46b5-921e-4f497810db31"
      },
      "execution_count": 94,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-94-e6ac3432fb76>:8: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
            "  history = model.fit_generator(train_data_generator,\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "31/31 [==============================] - 11s 93ms/step - loss: 36.3917 - accuracy: 0.0000e+00 - val_loss: 20.3412 - val_accuracy: 0.0000e+00\n",
            "Epoch 2/100\n",
            "31/31 [==============================] - 2s 75ms/step - loss: 19.9586 - accuracy: 0.0000e+00 - val_loss: 19.5849 - val_accuracy: 0.0000e+00\n",
            "Epoch 3/100\n",
            "31/31 [==============================] - 2s 58ms/step - loss: 19.0636 - accuracy: 0.0000e+00 - val_loss: 18.5204 - val_accuracy: 0.0000e+00\n",
            "Epoch 4/100\n",
            "31/31 [==============================] - 1s 43ms/step - loss: 18.3458 - accuracy: 0.0000e+00 - val_loss: 18.0469 - val_accuracy: 0.0000e+00\n",
            "Epoch 5/100\n",
            "31/31 [==============================] - 1s 42ms/step - loss: 18.0499 - accuracy: 0.0000e+00 - val_loss: 17.9226 - val_accuracy: 0.0000e+00\n",
            "Epoch 6/100\n",
            "31/31 [==============================] - 1s 43ms/step - loss: 17.9575 - accuracy: 0.0000e+00 - val_loss: 17.8717 - val_accuracy: 0.0000e+00\n",
            "Epoch 7/100\n",
            "31/31 [==============================] - 1s 42ms/step - loss: 17.9358 - accuracy: 0.0000e+00 - val_loss: 17.8932 - val_accuracy: 0.0000e+00\n",
            "Epoch 8/100\n",
            "31/31 [==============================] - 1s 43ms/step - loss: 17.9116 - accuracy: 0.0000e+00 - val_loss: 17.8258 - val_accuracy: 0.0000e+00\n",
            "Epoch 9/100\n",
            "31/31 [==============================] - 2s 65ms/step - loss: 17.8941 - accuracy: 0.0000e+00 - val_loss: 17.8628 - val_accuracy: 0.0000e+00\n",
            "Epoch 10/100\n",
            "31/31 [==============================] - 2s 71ms/step - loss: 17.8817 - accuracy: 0.0000e+00 - val_loss: 17.8415 - val_accuracy: 0.0000e+00\n",
            "Epoch 11/100\n",
            "31/31 [==============================] - 1s 45ms/step - loss: 17.8665 - accuracy: 0.0000e+00 - val_loss: 17.8562 - val_accuracy: 0.0000e+00\n",
            "Epoch 12/100\n",
            "31/31 [==============================] - 1s 42ms/step - loss: 17.8447 - accuracy: 0.0000e+00 - val_loss: 17.8276 - val_accuracy: 0.0000e+00\n",
            "Epoch 13/100\n",
            "31/31 [==============================] - 1s 44ms/step - loss: 17.8145 - accuracy: 0.0000e+00 - val_loss: 17.8316 - val_accuracy: 0.0000e+00\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.save(f'/content/drive/MyDrive/ML/model_for_samples.h5')"
      ],
      "metadata": {
        "id": "992iZQTAHGMN"
      },
      "execution_count": 83,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# optimzier와 learing rate별로 그래프 그리기\n",
        "'''\n",
        "optimzier 종류:\n",
        "SGD\n",
        "RMSprop\n",
        "Adam\n",
        "AdamW => X\n",
        "Adadelta\n",
        "Adagrad\n",
        "Adamax\n",
        "Adafactor => X\n",
        "Nadam\n",
        "Ftrl\n",
        "'''\n",
        "\n",
        "# 디렉토리 주소: 5_captcha or 4_captcha.\n",
        "\n",
        "def lr_and_opt_test(lr_list, opt_list):\n",
        "  for lr in lr_list:\n",
        "    for opt in opt_list:\n",
        "      print('****************** optimizer:',opt,'lr:',lr,'******************')\n",
        "      model = build_model(opt=opt)\n",
        "      # Add early stopping\n",
        "      es = keras.callbacks.EarlyStopping(monitor='val_loss',\n",
        "                                        patience=5,\n",
        "                                        restore_best_weights=True)\n",
        "\n",
        "      # Train the model\n",
        "      history = model.fit_generator(train_data_generator,\n",
        "                          validation_data=valid_data_generator,\n",
        "                          epochs=100,\n",
        "                          callbacks=[es])\n",
        "      \n",
        "      final_acc = history.history['accuracy'][-1]\n",
        "      final_val_acc = history.history['val_accuracy'][-1]\n",
        "\n",
        "      # 정확도 그래프 그리기 및 저장\n",
        "      plt.plot(history.history['accuracy'], color='red')\n",
        "      plt.plot(history.history['val_accuracy'], color='blue')\n",
        "      plt.title('model accuracy')\n",
        "      plt.ylabel('accuracy')\n",
        "      plt.xlabel('epoch')\n",
        "      plt.legend(['train_accuracy', 'test_accuracy'], loc='upper left')\n",
        "      plt.savefig(f'/content/drive/MyDrive/ML/{captcha}_captcha/{opt}/model_acc_{lr}.png') # 디렉토리 주소: 5_captcha or 4_captcha.\n",
        "\n",
        "      plt.clf() # 그래프 초기화\n",
        "\n",
        "      # loss 그래프 그리기 및 저장\n",
        "      plt.plot(history.history['loss'], color='brown')\n",
        "      plt.plot(history.history['val_loss'], color='green')\n",
        "      plt.title('model loss')\n",
        "      plt.ylabel('loss')\n",
        "      plt.xlabel('epoch')\n",
        "      plt.legend(['train_loss', 'test_loss'], loc='upper left')\n",
        "      plt.savefig(f'/content/drive/MyDrive/ML/{captcha}_captcha/{opt}/model_loss_{lr}.png') # 디렉토리 주소: 5_captcha or 4_captcha.\n",
        "\n",
        "      plt.clf() # 그래프 초기화\n",
        "\n",
        "      # acc과 loss 둘 다 반영한 그래프\n",
        "      plt.plot(history.history['accuracy'], color='red')\n",
        "      plt.plot(history.history['loss'], color='brown')\n",
        "      plt.plot(history.history['val_accuracy'], color='blue')\n",
        "      plt.plot(history.history['val_loss'], color='green')\n",
        "      plt.xlabel('epoch')\n",
        "      plt.legend(['train_accuracy', 'train_loss', 'test_accuracy', 'test_loss'], loc='upper left')\n",
        "      plt.savefig(f'/content/drive/MyDrive/ML/{captcha}_captcha/{opt}/model_acc_loss_{lr}.png') # 디렉토리 주소: 5_captcha or 4_captcha.\n",
        "\n",
        "      plt.clf() # 그래프 초기화\n",
        "\n",
        "      # log.txt에 값 저장하기. columns: opt,\tlr,\ttrain_acc,\ttest_acc\n",
        "      file_path = f'/content/drive/MyDrive/ML/{captcha}_captcha/log.txt' # 디렉토리 주소: 5_captcha or 4_captcha.\n",
        "      new_text_content = ''\n",
        "\n",
        "      with open(file_path,'r') as f:\n",
        "        lines = f.readlines()\n",
        "        for i, l in enumerate(lines):\n",
        "          new_string = l.strip()\n",
        "          if new_string:\n",
        "            new_text_content += new_string + '\\n'\n",
        "            \n",
        "      new_text_content += f'{opt}\\t{lr}\\t{final_acc}\\t{final_val_acc}' + '\\n'\n",
        "        \n",
        "      with open(file_path, 'w') as f:\n",
        "          f.write(new_text_content)\n",
        "\n",
        "      model.save(f'/content/drive/MyDrive/ML/{captcha}_captcha/Model/model_{opt}_{lr}.h5')\n",
        "\n",
        "lr_list = [0.002, 0.003]\n",
        "opt_list = ['SGD','Adam']\n",
        "lr_and_opt_test(lr_list, opt_list)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "K9eEwtijM20p",
        "outputId": "256773ea-4b20-4f07-e539-adba4da644e1"
      },
      "execution_count": 63,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "****************** optimizer: SGD lr: 0.002 ******************\n",
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-63-96bb6cf6ce3b>:29: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
            "  history = model.fit_generator(train_data_generator,\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "30/30 [==============================] - 18s 159ms/step - loss: 28.6261 - accuracy: 0.0000e+00 - val_loss: 16.5584 - val_accuracy: 0.0000e+00\n",
            "Epoch 2/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 16.3519 - accuracy: 0.0000e+00 - val_loss: 16.4885 - val_accuracy: 0.0000e+00\n",
            "Epoch 3/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.3359 - accuracy: 0.0000e+00 - val_loss: 16.4759 - val_accuracy: 0.0000e+00\n",
            "Epoch 4/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.3103 - accuracy: 0.0000e+00 - val_loss: 16.4025 - val_accuracy: 0.0000e+00\n",
            "Epoch 5/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 16.2920 - accuracy: 0.0000e+00 - val_loss: 16.3912 - val_accuracy: 0.0000e+00\n",
            "Epoch 6/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.2504 - accuracy: 0.0000e+00 - val_loss: 16.4101 - val_accuracy: 0.0000e+00\n",
            "Epoch 7/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 16.1696 - accuracy: 0.0000e+00 - val_loss: 16.1527 - val_accuracy: 0.0000e+00\n",
            "Epoch 8/100\n",
            "30/30 [==============================] - 2s 53ms/step - loss: 16.0873 - accuracy: 0.0000e+00 - val_loss: 16.1583 - val_accuracy: 0.0000e+00\n",
            "Epoch 9/100\n",
            "30/30 [==============================] - 2s 62ms/step - loss: 16.0463 - accuracy: 0.0000e+00 - val_loss: 16.0251 - val_accuracy: 0.0000e+00\n",
            "Epoch 10/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.0135 - accuracy: 0.0000e+00 - val_loss: 15.9206 - val_accuracy: 0.0000e+00\n",
            "Epoch 11/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.9862 - accuracy: 0.0000e+00 - val_loss: 15.9115 - val_accuracy: 0.0000e+00\n",
            "Epoch 12/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.9352 - accuracy: 0.0000e+00 - val_loss: 15.8473 - val_accuracy: 0.0000e+00\n",
            "Epoch 13/100\n",
            "30/30 [==============================] - 1s 39ms/step - loss: 15.8259 - accuracy: 0.0000e+00 - val_loss: 15.6126 - val_accuracy: 0.0000e+00\n",
            "Epoch 14/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 15.6197 - accuracy: 0.0000e+00 - val_loss: 15.4298 - val_accuracy: 0.0000e+00\n",
            "Epoch 15/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.4996 - accuracy: 0.0000e+00 - val_loss: 15.4445 - val_accuracy: 0.0000e+00\n",
            "Epoch 16/100\n",
            "30/30 [==============================] - 2s 60ms/step - loss: 15.4405 - accuracy: 0.0000e+00 - val_loss: 15.4179 - val_accuracy: 0.0000e+00\n",
            "Epoch 17/100\n",
            "30/30 [==============================] - 2s 75ms/step - loss: 15.3801 - accuracy: 0.0000e+00 - val_loss: 15.3504 - val_accuracy: 0.0000e+00\n",
            "Epoch 18/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 15.3286 - accuracy: 0.0000e+00 - val_loss: 15.3532 - val_accuracy: 0.0000e+00\n",
            "Epoch 19/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.2851 - accuracy: 0.0000e+00 - val_loss: 15.2366 - val_accuracy: 0.0000e+00\n",
            "Epoch 20/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.2237 - accuracy: 0.0000e+00 - val_loss: 15.2001 - val_accuracy: 0.0000e+00\n",
            "Epoch 21/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.1584 - accuracy: 0.0000e+00 - val_loss: 15.1644 - val_accuracy: 0.0000e+00\n",
            "Epoch 22/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.0792 - accuracy: 0.0000e+00 - val_loss: 15.0497 - val_accuracy: 0.0000e+00\n",
            "Epoch 23/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 14.9860 - accuracy: 0.0000e+00 - val_loss: 15.0200 - val_accuracy: 0.0000e+00\n",
            "Epoch 24/100\n",
            "30/30 [==============================] - 1s 47ms/step - loss: 14.8966 - accuracy: 0.0000e+00 - val_loss: 14.8365 - val_accuracy: 0.0000e+00\n",
            "Epoch 25/100\n",
            "30/30 [==============================] - 2s 68ms/step - loss: 14.7541 - accuracy: 0.0000e+00 - val_loss: 14.6640 - val_accuracy: 0.0000e+00\n",
            "Epoch 26/100\n",
            "30/30 [==============================] - 2s 60ms/step - loss: 14.6053 - accuracy: 0.0000e+00 - val_loss: 14.4026 - val_accuracy: 0.0000e+00\n",
            "Epoch 27/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 14.4560 - accuracy: 0.0000e+00 - val_loss: 14.3192 - val_accuracy: 0.0000e+00\n",
            "Epoch 28/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 14.3259 - accuracy: 0.0000e+00 - val_loss: 14.3823 - val_accuracy: 0.0000e+00\n",
            "Epoch 29/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 14.1538 - accuracy: 0.0000e+00 - val_loss: 13.9436 - val_accuracy: 0.0000e+00\n",
            "Epoch 30/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 13.9440 - accuracy: 0.0000e+00 - val_loss: 13.6656 - val_accuracy: 0.0000e+00\n",
            "Epoch 31/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 13.7436 - accuracy: 0.0000e+00 - val_loss: 13.5423 - val_accuracy: 0.0000e+00\n",
            "Epoch 32/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 13.5855 - accuracy: 0.0000e+00 - val_loss: 13.3372 - val_accuracy: 0.0000e+00\n",
            "Epoch 33/100\n",
            "30/30 [==============================] - 2s 61ms/step - loss: 13.4343 - accuracy: 0.0000e+00 - val_loss: 13.2959 - val_accuracy: 0.0000e+00\n",
            "Epoch 34/100\n",
            "30/30 [==============================] - 2s 72ms/step - loss: 13.2657 - accuracy: 0.0000e+00 - val_loss: 13.1726 - val_accuracy: 0.0000e+00\n",
            "Epoch 35/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 13.1261 - accuracy: 0.0000e+00 - val_loss: 12.9533 - val_accuracy: 0.0000e+00\n",
            "Epoch 36/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 12.9491 - accuracy: 0.0000e+00 - val_loss: 12.8778 - val_accuracy: 0.0000e+00\n",
            "Epoch 37/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 12.8480 - accuracy: 0.0000e+00 - val_loss: 12.7578 - val_accuracy: 0.0000e+00\n",
            "Epoch 38/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 12.6588 - accuracy: 0.0000e+00 - val_loss: 12.7540 - val_accuracy: 0.0000e+00\n",
            "Epoch 39/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 12.5430 - accuracy: 0.0000e+00 - val_loss: 12.4964 - val_accuracy: 0.0000e+00\n",
            "Epoch 40/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 12.3673 - accuracy: 0.0000e+00 - val_loss: 12.3229 - val_accuracy: 0.0000e+00\n",
            "Epoch 41/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 12.2259 - accuracy: 0.0000e+00 - val_loss: 12.3030 - val_accuracy: 0.0000e+00\n",
            "Epoch 42/100\n",
            "30/30 [==============================] - 2s 52ms/step - loss: 12.0221 - accuracy: 0.0000e+00 - val_loss: 11.9321 - val_accuracy: 0.0000e+00\n",
            "Epoch 43/100\n",
            "30/30 [==============================] - 2s 72ms/step - loss: 11.7002 - accuracy: 0.0000e+00 - val_loss: 11.5551 - val_accuracy: 0.0000e+00\n",
            "Epoch 44/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 11.3163 - accuracy: 0.0000e+00 - val_loss: 11.1832 - val_accuracy: 0.0000e+00\n",
            "Epoch 45/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 10.8686 - accuracy: 0.0000e+00 - val_loss: 10.7052 - val_accuracy: 0.0000e+00\n",
            "Epoch 46/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 10.3562 - accuracy: 0.0000e+00 - val_loss: 10.0901 - val_accuracy: 0.0000e+00\n",
            "Epoch 47/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 9.9115 - accuracy: 0.0000e+00 - val_loss: 9.5783 - val_accuracy: 0.0000e+00\n",
            "Epoch 48/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 9.4413 - accuracy: 0.0000e+00 - val_loss: 9.3665 - val_accuracy: 0.0000e+00\n",
            "Epoch 49/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 8.9620 - accuracy: 0.0000e+00 - val_loss: 8.6895 - val_accuracy: 0.0000e+00\n",
            "Epoch 50/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 8.4977 - accuracy: 0.0000e+00 - val_loss: 8.0493 - val_accuracy: 0.0000e+00\n",
            "Epoch 51/100\n",
            "30/30 [==============================] - 2s 52ms/step - loss: 7.9842 - accuracy: 0.0000e+00 - val_loss: 7.5837 - val_accuracy: 0.0000e+00\n",
            "Epoch 52/100\n",
            "30/30 [==============================] - 2s 69ms/step - loss: 7.5368 - accuracy: 0.0000e+00 - val_loss: 7.0528 - val_accuracy: 0.0000e+00\n",
            "Epoch 53/100\n",
            "30/30 [==============================] - 2s 55ms/step - loss: 7.0654 - accuracy: 0.0000e+00 - val_loss: 6.4594 - val_accuracy: 0.0000e+00\n",
            "Epoch 54/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 6.5891 - accuracy: 0.0000e+00 - val_loss: 6.0676 - val_accuracy: 0.0000e+00\n",
            "Epoch 55/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 6.0714 - accuracy: 0.0000e+00 - val_loss: 5.5243 - val_accuracy: 0.0000e+00\n",
            "Epoch 56/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 5.5769 - accuracy: 0.0000e+00 - val_loss: 4.9767 - val_accuracy: 0.0000e+00\n",
            "Epoch 57/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 5.0309 - accuracy: 0.0000e+00 - val_loss: 4.5565 - val_accuracy: 0.0000e+00\n",
            "Epoch 58/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 4.5788 - accuracy: 0.0000e+00 - val_loss: 3.8139 - val_accuracy: 0.0000e+00\n",
            "Epoch 59/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 4.0748 - accuracy: 0.0000e+00 - val_loss: 3.2470 - val_accuracy: 0.0000e+00\n",
            "Epoch 60/100\n",
            "30/30 [==============================] - 2s 65ms/step - loss: 3.5597 - accuracy: 0.0000e+00 - val_loss: 2.8687 - val_accuracy: 0.0000e+00\n",
            "Epoch 61/100\n",
            "30/30 [==============================] - 2s 60ms/step - loss: 3.1995 - accuracy: 0.0000e+00 - val_loss: 2.5030 - val_accuracy: 0.0000e+00\n",
            "Epoch 62/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 2.7659 - accuracy: 0.0011 - val_loss: 2.0918 - val_accuracy: 0.0288\n",
            "Epoch 63/100\n",
            "30/30 [==============================] - 1s 47ms/step - loss: 2.4870 - accuracy: 0.0000e+00 - val_loss: 1.8333 - val_accuracy: 0.1058\n",
            "Epoch 64/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 2.1121 - accuracy: 0.0160 - val_loss: 1.4942 - val_accuracy: 0.2019\n",
            "Epoch 65/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 1.8355 - accuracy: 0.0641 - val_loss: 1.3201 - val_accuracy: 0.3558\n",
            "Epoch 66/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 1.6505 - accuracy: 0.1165 - val_loss: 1.3108 - val_accuracy: 0.4615\n",
            "Epoch 67/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 1.4709 - accuracy: 0.1987 - val_loss: 1.0746 - val_accuracy: 0.4808\n",
            "Epoch 68/100\n",
            "30/30 [==============================] - 2s 74ms/step - loss: 1.2733 - accuracy: 0.2788 - val_loss: 1.1059 - val_accuracy: 0.5577\n",
            "Epoch 69/100\n",
            "30/30 [==============================] - 2s 68ms/step - loss: 1.1141 - accuracy: 0.3397 - val_loss: 1.0444 - val_accuracy: 0.5962\n",
            "Epoch 70/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 1.0754 - accuracy: 0.4060 - val_loss: 0.9645 - val_accuracy: 0.6250\n",
            "Epoch 71/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.9848 - accuracy: 0.4573 - val_loss: 0.8201 - val_accuracy: 0.6442\n",
            "Epoch 72/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.9292 - accuracy: 0.4925 - val_loss: 0.7899 - val_accuracy: 0.6635\n",
            "Epoch 73/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 0.8283 - accuracy: 0.5673 - val_loss: 0.9745 - val_accuracy: 0.6442\n",
            "Epoch 74/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 0.8304 - accuracy: 0.5491 - val_loss: 0.8405 - val_accuracy: 0.6731\n",
            "Epoch 75/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 0.7643 - accuracy: 0.5962 - val_loss: 0.6775 - val_accuracy: 0.7115\n",
            "Epoch 76/100\n",
            "30/30 [==============================] - 2s 60ms/step - loss: 0.6765 - accuracy: 0.6378 - val_loss: 0.6050 - val_accuracy: 0.7404\n",
            "Epoch 77/100\n",
            "30/30 [==============================] - 2s 63ms/step - loss: 0.6390 - accuracy: 0.6581 - val_loss: 0.5881 - val_accuracy: 0.7596\n",
            "Epoch 78/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.6224 - accuracy: 0.6838 - val_loss: 0.7024 - val_accuracy: 0.7596\n",
            "Epoch 79/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.5597 - accuracy: 0.7190 - val_loss: 0.7059 - val_accuracy: 0.7308\n",
            "Epoch 80/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.5731 - accuracy: 0.7244 - val_loss: 0.5306 - val_accuracy: 0.7500\n",
            "Epoch 81/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 0.4940 - accuracy: 0.7479 - val_loss: 0.4374 - val_accuracy: 0.7885\n",
            "Epoch 82/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 0.4876 - accuracy: 0.7746 - val_loss: 0.5627 - val_accuracy: 0.7788\n",
            "Epoch 83/100\n",
            "30/30 [==============================] - 2s 69ms/step - loss: 0.4280 - accuracy: 0.7756 - val_loss: 0.5369 - val_accuracy: 0.7981\n",
            "Epoch 84/100\n",
            "30/30 [==============================] - 2s 55ms/step - loss: 0.4294 - accuracy: 0.7799 - val_loss: 0.4645 - val_accuracy: 0.8269\n",
            "Epoch 85/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 0.4196 - accuracy: 0.8013 - val_loss: 0.5631 - val_accuracy: 0.8077\n",
            "Epoch 86/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.4043 - accuracy: 0.8301 - val_loss: 0.4102 - val_accuracy: 0.8558\n",
            "Epoch 87/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.3915 - accuracy: 0.8194 - val_loss: 0.4245 - val_accuracy: 0.8846\n",
            "Epoch 88/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.3373 - accuracy: 0.8429 - val_loss: 0.6256 - val_accuracy: 0.8365\n",
            "Epoch 89/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.3014 - accuracy: 0.8600 - val_loss: 0.5180 - val_accuracy: 0.8750\n",
            "Epoch 90/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.3393 - accuracy: 0.8739 - val_loss: 0.4124 - val_accuracy: 0.8750\n",
            "Epoch 91/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.2989 - accuracy: 0.8739 - val_loss: 0.4634 - val_accuracy: 0.8365\n",
            "****************** optimizer: Adam lr: 0.002 ******************\n",
            "Epoch 1/100\n",
            "30/30 [==============================] - 11s 97ms/step - loss: 21.6820 - accuracy: 0.0000e+00 - val_loss: 16.5166 - val_accuracy: 0.0000e+00\n",
            "Epoch 2/100\n",
            "30/30 [==============================] - 2s 77ms/step - loss: 16.3341 - accuracy: 0.0000e+00 - val_loss: 16.3616 - val_accuracy: 0.0000e+00\n",
            "Epoch 3/100\n",
            "30/30 [==============================] - 2s 52ms/step - loss: 16.3119 - accuracy: 0.0000e+00 - val_loss: 16.3617 - val_accuracy: 0.0000e+00\n",
            "Epoch 4/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 16.3070 - accuracy: 0.0000e+00 - val_loss: 16.3718 - val_accuracy: 0.0000e+00\n",
            "Epoch 5/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.3104 - accuracy: 0.0000e+00 - val_loss: 16.3588 - val_accuracy: 0.0000e+00\n",
            "Epoch 6/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 16.3109 - accuracy: 0.0000e+00 - val_loss: 16.3803 - val_accuracy: 0.0000e+00\n",
            "Epoch 7/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.3129 - accuracy: 0.0000e+00 - val_loss: 16.3639 - val_accuracy: 0.0000e+00\n",
            "Epoch 8/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 16.3104 - accuracy: 0.0000e+00 - val_loss: 16.3658 - val_accuracy: 0.0000e+00\n",
            "Epoch 9/100\n",
            "30/30 [==============================] - 2s 62ms/step - loss: 16.3050 - accuracy: 0.0000e+00 - val_loss: 16.3480 - val_accuracy: 0.0000e+00\n",
            "Epoch 10/100\n",
            "30/30 [==============================] - 2s 75ms/step - loss: 16.2871 - accuracy: 0.0000e+00 - val_loss: 16.3698 - val_accuracy: 0.0000e+00\n",
            "Epoch 11/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 16.2751 - accuracy: 0.0000e+00 - val_loss: 16.3568 - val_accuracy: 0.0000e+00\n",
            "Epoch 12/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 16.2544 - accuracy: 0.0000e+00 - val_loss: 16.2667 - val_accuracy: 0.0000e+00\n",
            "Epoch 13/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.2272 - accuracy: 0.0000e+00 - val_loss: 16.1908 - val_accuracy: 0.0000e+00\n",
            "Epoch 14/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 16.1728 - accuracy: 0.0000e+00 - val_loss: 16.1016 - val_accuracy: 0.0000e+00\n",
            "Epoch 15/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.0887 - accuracy: 0.0000e+00 - val_loss: 16.0281 - val_accuracy: 0.0000e+00\n",
            "Epoch 16/100\n",
            "30/30 [==============================] - 2s 55ms/step - loss: 16.0464 - accuracy: 0.0000e+00 - val_loss: 15.9416 - val_accuracy: 0.0000e+00\n",
            "Epoch 17/100\n",
            "30/30 [==============================] - 2s 63ms/step - loss: 16.0024 - accuracy: 0.0000e+00 - val_loss: 15.9080 - val_accuracy: 0.0000e+00\n",
            "Epoch 18/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.9664 - accuracy: 0.0000e+00 - val_loss: 15.8870 - val_accuracy: 0.0000e+00\n",
            "Epoch 19/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.9300 - accuracy: 0.0000e+00 - val_loss: 15.8449 - val_accuracy: 0.0000e+00\n",
            "Epoch 20/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 15.9179 - accuracy: 0.0000e+00 - val_loss: 15.8208 - val_accuracy: 0.0000e+00\n",
            "Epoch 21/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.8993 - accuracy: 0.0000e+00 - val_loss: 15.7905 - val_accuracy: 0.0000e+00\n",
            "Epoch 22/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 15.8925 - accuracy: 0.0000e+00 - val_loss: 15.7734 - val_accuracy: 0.0000e+00\n",
            "Epoch 23/100\n",
            "30/30 [==============================] - 1s 48ms/step - loss: 15.8683 - accuracy: 0.0000e+00 - val_loss: 15.7506 - val_accuracy: 0.0000e+00\n",
            "Epoch 24/100\n",
            "30/30 [==============================] - 2s 64ms/step - loss: 15.8533 - accuracy: 0.0000e+00 - val_loss: 15.7457 - val_accuracy: 0.0000e+00\n",
            "Epoch 25/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 15.8440 - accuracy: 0.0000e+00 - val_loss: 15.7562 - val_accuracy: 0.0000e+00\n",
            "Epoch 26/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.8308 - accuracy: 0.0000e+00 - val_loss: 15.7463 - val_accuracy: 0.0000e+00\n",
            "Epoch 27/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.8359 - accuracy: 0.0000e+00 - val_loss: 15.7412 - val_accuracy: 0.0000e+00\n",
            "Epoch 28/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.8095 - accuracy: 0.0000e+00 - val_loss: 15.7280 - val_accuracy: 0.0000e+00\n",
            "Epoch 29/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.8028 - accuracy: 0.0000e+00 - val_loss: 15.7232 - val_accuracy: 0.0000e+00\n",
            "Epoch 30/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.8003 - accuracy: 0.0000e+00 - val_loss: 15.7148 - val_accuracy: 0.0000e+00\n",
            "Epoch 31/100\n",
            "30/30 [==============================] - 2s 73ms/step - loss: 15.7931 - accuracy: 0.0000e+00 - val_loss: 15.7316 - val_accuracy: 0.0000e+00\n",
            "Epoch 32/100\n",
            "30/30 [==============================] - 2s 50ms/step - loss: 15.7791 - accuracy: 0.0000e+00 - val_loss: 15.7009 - val_accuracy: 0.0000e+00\n",
            "Epoch 33/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.7965 - accuracy: 0.0000e+00 - val_loss: 15.7521 - val_accuracy: 0.0000e+00\n",
            "Epoch 34/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 15.7761 - accuracy: 0.0000e+00 - val_loss: 15.7371 - val_accuracy: 0.0000e+00\n",
            "Epoch 35/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.7630 - accuracy: 0.0000e+00 - val_loss: 15.6936 - val_accuracy: 0.0000e+00\n",
            "Epoch 36/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.7563 - accuracy: 0.0000e+00 - val_loss: 15.7210 - val_accuracy: 0.0000e+00\n",
            "Epoch 37/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.7509 - accuracy: 0.0000e+00 - val_loss: 15.7254 - val_accuracy: 0.0000e+00\n",
            "Epoch 38/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 15.7557 - accuracy: 0.0000e+00 - val_loss: 15.7385 - val_accuracy: 0.0000e+00\n",
            "Epoch 39/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 15.7289 - accuracy: 0.0000e+00 - val_loss: 15.7153 - val_accuracy: 0.0000e+00\n",
            "Epoch 40/100\n",
            "30/30 [==============================] - 2s 66ms/step - loss: 15.7401 - accuracy: 0.0000e+00 - val_loss: 15.7023 - val_accuracy: 0.0000e+00\n",
            "****************** optimizer: SGD lr: 0.003 ******************\n",
            "Epoch 1/100\n",
            "30/30 [==============================] - 10s 123ms/step - loss: 36.2605 - accuracy: 0.0000e+00 - val_loss: 16.4999 - val_accuracy: 0.0000e+00\n",
            "Epoch 2/100\n",
            "30/30 [==============================] - 2s 65ms/step - loss: 16.3483 - accuracy: 0.0000e+00 - val_loss: 16.4144 - val_accuracy: 0.0000e+00\n",
            "Epoch 3/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.3043 - accuracy: 0.0000e+00 - val_loss: 16.4399 - val_accuracy: 0.0000e+00\n",
            "Epoch 4/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.2254 - accuracy: 0.0000e+00 - val_loss: 16.1387 - val_accuracy: 0.0000e+00\n",
            "Epoch 5/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.0942 - accuracy: 0.0000e+00 - val_loss: 16.0281 - val_accuracy: 0.0000e+00\n",
            "Epoch 6/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.0241 - accuracy: 0.0000e+00 - val_loss: 15.9035 - val_accuracy: 0.0000e+00\n",
            "Epoch 7/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.9850 - accuracy: 0.0000e+00 - val_loss: 15.9285 - val_accuracy: 0.0000e+00\n",
            "Epoch 8/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.9748 - accuracy: 0.0000e+00 - val_loss: 16.0051 - val_accuracy: 0.0000e+00\n",
            "Epoch 9/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.9412 - accuracy: 0.0000e+00 - val_loss: 15.8165 - val_accuracy: 0.0000e+00\n",
            "Epoch 10/100\n",
            "30/30 [==============================] - 2s 67ms/step - loss: 15.7876 - accuracy: 0.0000e+00 - val_loss: 15.5171 - val_accuracy: 0.0000e+00\n",
            "Epoch 11/100\n",
            "30/30 [==============================] - 2s 58ms/step - loss: 15.5581 - accuracy: 0.0000e+00 - val_loss: 15.3387 - val_accuracy: 0.0000e+00\n",
            "Epoch 12/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.4216 - accuracy: 0.0000e+00 - val_loss: 15.3664 - val_accuracy: 0.0000e+00\n",
            "Epoch 13/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.3078 - accuracy: 0.0000e+00 - val_loss: 15.2634 - val_accuracy: 0.0000e+00\n",
            "Epoch 14/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 15.1854 - accuracy: 0.0000e+00 - val_loss: 15.0388 - val_accuracy: 0.0000e+00\n",
            "Epoch 15/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 14.9505 - accuracy: 0.0000e+00 - val_loss: 14.8056 - val_accuracy: 0.0000e+00\n",
            "Epoch 16/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 14.6832 - accuracy: 0.0000e+00 - val_loss: 14.5305 - val_accuracy: 0.0000e+00\n",
            "Epoch 17/100\n",
            "30/30 [==============================] - 2s 57ms/step - loss: 14.4615 - accuracy: 0.0000e+00 - val_loss: 14.3017 - val_accuracy: 0.0000e+00\n",
            "Epoch 18/100\n",
            "30/30 [==============================] - 2s 76ms/step - loss: 14.3227 - accuracy: 0.0000e+00 - val_loss: 14.1833 - val_accuracy: 0.0000e+00\n",
            "Epoch 19/100\n",
            "30/30 [==============================] - 2s 50ms/step - loss: 14.1794 - accuracy: 0.0000e+00 - val_loss: 14.1003 - val_accuracy: 0.0000e+00\n",
            "Epoch 20/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 14.0542 - accuracy: 0.0000e+00 - val_loss: 13.8969 - val_accuracy: 0.0000e+00\n",
            "Epoch 21/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 13.9522 - accuracy: 0.0000e+00 - val_loss: 13.8311 - val_accuracy: 0.0000e+00\n",
            "Epoch 22/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 13.7872 - accuracy: 0.0000e+00 - val_loss: 13.7603 - val_accuracy: 0.0000e+00\n",
            "Epoch 23/100\n",
            "30/30 [==============================] - 2s 51ms/step - loss: 13.6888 - accuracy: 0.0000e+00 - val_loss: 13.6821 - val_accuracy: 0.0000e+00\n",
            "Epoch 24/100\n",
            "30/30 [==============================] - 2s 71ms/step - loss: 13.5131 - accuracy: 0.0000e+00 - val_loss: 13.3755 - val_accuracy: 0.0000e+00\n",
            "Epoch 25/100\n",
            "30/30 [==============================] - 2s 58ms/step - loss: 13.3831 - accuracy: 0.0000e+00 - val_loss: 13.4959 - val_accuracy: 0.0000e+00\n",
            "Epoch 26/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 13.2755 - accuracy: 0.0000e+00 - val_loss: 13.3060 - val_accuracy: 0.0000e+00\n",
            "Epoch 27/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 13.1590 - accuracy: 0.0000e+00 - val_loss: 13.0721 - val_accuracy: 0.0000e+00\n",
            "Epoch 28/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 13.0005 - accuracy: 0.0000e+00 - val_loss: 12.9304 - val_accuracy: 0.0000e+00\n",
            "Epoch 29/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 12.8908 - accuracy: 0.0000e+00 - val_loss: 12.9351 - val_accuracy: 0.0000e+00\n",
            "Epoch 30/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 12.7514 - accuracy: 0.0000e+00 - val_loss: 12.7141 - val_accuracy: 0.0000e+00\n",
            "Epoch 31/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 12.5783 - accuracy: 0.0000e+00 - val_loss: 12.5886 - val_accuracy: 0.0000e+00\n",
            "Epoch 32/100\n",
            "30/30 [==============================] - 2s 63ms/step - loss: 12.4732 - accuracy: 0.0000e+00 - val_loss: 12.4573 - val_accuracy: 0.0000e+00\n",
            "Epoch 33/100\n",
            "30/30 [==============================] - 2s 71ms/step - loss: 12.2471 - accuracy: 0.0000e+00 - val_loss: 12.2870 - val_accuracy: 0.0000e+00\n",
            "Epoch 34/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 11.9703 - accuracy: 0.0000e+00 - val_loss: 12.0825 - val_accuracy: 0.0000e+00\n",
            "Epoch 35/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 11.6489 - accuracy: 0.0000e+00 - val_loss: 11.5018 - val_accuracy: 0.0000e+00\n",
            "Epoch 36/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 11.2137 - accuracy: 0.0000e+00 - val_loss: 11.3216 - val_accuracy: 0.0000e+00\n",
            "Epoch 37/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 10.8517 - accuracy: 0.0000e+00 - val_loss: 10.7961 - val_accuracy: 0.0000e+00\n",
            "Epoch 38/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 10.4260 - accuracy: 0.0000e+00 - val_loss: 10.3184 - val_accuracy: 0.0000e+00\n",
            "Epoch 39/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 9.9195 - accuracy: 0.0000e+00 - val_loss: 9.8216 - val_accuracy: 0.0000e+00\n",
            "Epoch 40/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 9.4447 - accuracy: 0.0000e+00 - val_loss: 9.2032 - val_accuracy: 0.0000e+00\n",
            "Epoch 41/100\n",
            "30/30 [==============================] - 2s 57ms/step - loss: 8.9303 - accuracy: 0.0000e+00 - val_loss: 8.6301 - val_accuracy: 0.0000e+00\n",
            "Epoch 42/100\n",
            "30/30 [==============================] - 2s 69ms/step - loss: 8.4701 - accuracy: 0.0000e+00 - val_loss: 8.0637 - val_accuracy: 0.0000e+00\n",
            "Epoch 43/100\n",
            "30/30 [==============================] - 2s 50ms/step - loss: 7.9585 - accuracy: 0.0000e+00 - val_loss: 7.5482 - val_accuracy: 0.0000e+00\n",
            "Epoch 44/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 7.4261 - accuracy: 0.0000e+00 - val_loss: 7.0025 - val_accuracy: 0.0000e+00\n",
            "Epoch 45/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 6.9393 - accuracy: 0.0000e+00 - val_loss: 6.4496 - val_accuracy: 0.0000e+00\n",
            "Epoch 46/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 6.4543 - accuracy: 0.0000e+00 - val_loss: 5.8498 - val_accuracy: 0.0000e+00\n",
            "Epoch 47/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 5.9576 - accuracy: 0.0000e+00 - val_loss: 5.3323 - val_accuracy: 0.0000e+00\n",
            "Epoch 48/100\n",
            "30/30 [==============================] - 2s 65ms/step - loss: 5.4297 - accuracy: 0.0000e+00 - val_loss: 4.8080 - val_accuracy: 0.0000e+00\n",
            "Epoch 49/100\n",
            "30/30 [==============================] - 2s 68ms/step - loss: 4.9483 - accuracy: 0.0000e+00 - val_loss: 4.1653 - val_accuracy: 0.0000e+00\n",
            "Epoch 50/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 4.4928 - accuracy: 0.0000e+00 - val_loss: 3.7803 - val_accuracy: 0.0000e+00\n",
            "Epoch 51/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 3.9891 - accuracy: 0.0000e+00 - val_loss: 3.4466 - val_accuracy: 0.0000e+00\n",
            "Epoch 52/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 3.6419 - accuracy: 0.0000e+00 - val_loss: 2.8691 - val_accuracy: 0.0000e+00\n",
            "Epoch 53/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 3.2735 - accuracy: 0.0000e+00 - val_loss: 2.6491 - val_accuracy: 0.0000e+00\n",
            "Epoch 54/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 2.8906 - accuracy: 0.0000e+00 - val_loss: 2.3256 - val_accuracy: 0.0096\n",
            "Epoch 55/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 2.5745 - accuracy: 0.0011 - val_loss: 2.0325 - val_accuracy: 0.0481\n",
            "Epoch 56/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 2.3121 - accuracy: 0.0064 - val_loss: 1.6392 - val_accuracy: 0.1154\n",
            "Epoch 57/100\n",
            "30/30 [==============================] - 2s 79ms/step - loss: 2.0432 - accuracy: 0.0246 - val_loss: 1.4269 - val_accuracy: 0.1346\n",
            "Epoch 58/100\n",
            "30/30 [==============================] - 2s 52ms/step - loss: 1.7676 - accuracy: 0.0588 - val_loss: 1.4076 - val_accuracy: 0.3173\n",
            "Epoch 59/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 1.6428 - accuracy: 0.0887 - val_loss: 1.3756 - val_accuracy: 0.3654\n",
            "Epoch 60/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 1.4343 - accuracy: 0.1709 - val_loss: 1.1566 - val_accuracy: 0.4327\n",
            "Epoch 61/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 1.2527 - accuracy: 0.2553 - val_loss: 1.0051 - val_accuracy: 0.5769\n",
            "Epoch 62/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 1.1490 - accuracy: 0.3205 - val_loss: 0.9089 - val_accuracy: 0.6154\n",
            "Epoch 63/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 1.0118 - accuracy: 0.4038 - val_loss: 0.7966 - val_accuracy: 0.6346\n",
            "Epoch 64/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 0.9313 - accuracy: 0.4295 - val_loss: 0.8865 - val_accuracy: 0.6827\n",
            "Epoch 65/100\n",
            "30/30 [==============================] - 2s 75ms/step - loss: 0.8067 - accuracy: 0.5192 - val_loss: 0.7287 - val_accuracy: 0.7212\n",
            "Epoch 66/100\n",
            "30/30 [==============================] - 2s 59ms/step - loss: 0.7392 - accuracy: 0.5780 - val_loss: 0.5495 - val_accuracy: 0.7404\n",
            "Epoch 67/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 0.7176 - accuracy: 0.6026 - val_loss: 0.5244 - val_accuracy: 0.7981\n",
            "Epoch 68/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.6379 - accuracy: 0.6496 - val_loss: 0.5110 - val_accuracy: 0.7692\n",
            "Epoch 69/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.5272 - accuracy: 0.6880 - val_loss: 0.4640 - val_accuracy: 0.7596\n",
            "Epoch 70/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.5279 - accuracy: 0.7233 - val_loss: 0.3642 - val_accuracy: 0.8654\n",
            "Epoch 71/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 0.4762 - accuracy: 0.7468 - val_loss: 0.4154 - val_accuracy: 0.8462\n",
            "Epoch 72/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.4349 - accuracy: 0.7607 - val_loss: 0.3714 - val_accuracy: 0.8558\n",
            "Epoch 73/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.3968 - accuracy: 0.8066 - val_loss: 0.4513 - val_accuracy: 0.8654\n",
            "Epoch 74/100\n",
            "30/30 [==============================] - 2s 61ms/step - loss: 0.3820 - accuracy: 0.8141 - val_loss: 0.4077 - val_accuracy: 0.8558\n",
            "Epoch 75/100\n",
            "30/30 [==============================] - 2s 77ms/step - loss: 0.3677 - accuracy: 0.8162 - val_loss: 0.4097 - val_accuracy: 0.8846\n",
            "****************** optimizer: Adam lr: 0.003 ******************\n",
            "Epoch 1/100\n",
            "30/30 [==============================] - 13s 132ms/step - loss: 22.0055 - accuracy: 0.0000e+00 - val_loss: 16.5121 - val_accuracy: 0.0000e+00\n",
            "Epoch 2/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 16.3306 - accuracy: 0.0000e+00 - val_loss: 16.3668 - val_accuracy: 0.0000e+00\n",
            "Epoch 3/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 16.3149 - accuracy: 0.0000e+00 - val_loss: 16.3925 - val_accuracy: 0.0000e+00\n",
            "Epoch 4/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 16.3184 - accuracy: 0.0000e+00 - val_loss: 16.4561 - val_accuracy: 0.0000e+00\n",
            "Epoch 5/100\n",
            "30/30 [==============================] - 1s 40ms/step - loss: 16.3127 - accuracy: 0.0000e+00 - val_loss: 16.3740 - val_accuracy: 0.0000e+00\n",
            "Epoch 6/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 16.3215 - accuracy: 0.0000e+00 - val_loss: 16.3641 - val_accuracy: 0.0000e+00\n",
            "Epoch 7/100\n",
            "30/30 [==============================] - 2s 70ms/step - loss: 16.3096 - accuracy: 0.0000e+00 - val_loss: 16.3728 - val_accuracy: 0.0000e+00\n",
            "Epoch 8/100\n",
            "30/30 [==============================] - 2s 52ms/step - loss: 16.3086 - accuracy: 0.0000e+00 - val_loss: 16.3774 - val_accuracy: 0.0000e+00\n",
            "Epoch 9/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 16.2980 - accuracy: 0.0000e+00 - val_loss: 16.3828 - val_accuracy: 0.0000e+00\n",
            "Epoch 10/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 16.2930 - accuracy: 0.0000e+00 - val_loss: 16.3401 - val_accuracy: 0.0000e+00\n",
            "Epoch 11/100\n",
            "30/30 [==============================] - 1s 46ms/step - loss: 16.2764 - accuracy: 0.0000e+00 - val_loss: 16.3227 - val_accuracy: 0.0000e+00\n",
            "Epoch 12/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.2609 - accuracy: 0.0000e+00 - val_loss: 16.3288 - val_accuracy: 0.0000e+00\n",
            "Epoch 13/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 16.2442 - accuracy: 0.0000e+00 - val_loss: 16.2983 - val_accuracy: 0.0000e+00\n",
            "Epoch 14/100\n",
            "30/30 [==============================] - 2s 63ms/step - loss: 16.2081 - accuracy: 0.0000e+00 - val_loss: 16.2873 - val_accuracy: 0.0000e+00\n",
            "Epoch 15/100\n",
            "30/30 [==============================] - 2s 71ms/step - loss: 16.1391 - accuracy: 0.0000e+00 - val_loss: 16.1227 - val_accuracy: 0.0000e+00\n",
            "Epoch 16/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 15.9667 - accuracy: 0.0000e+00 - val_loss: 15.8413 - val_accuracy: 0.0000e+00\n",
            "Epoch 17/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 15.7477 - accuracy: 0.0000e+00 - val_loss: 15.7142 - val_accuracy: 0.0000e+00\n",
            "Epoch 18/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.4898 - accuracy: 0.0000e+00 - val_loss: 15.3327 - val_accuracy: 0.0000e+00\n",
            "Epoch 19/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 15.1863 - accuracy: 0.0000e+00 - val_loss: 14.9553 - val_accuracy: 0.0000e+00\n",
            "Epoch 20/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 14.6569 - accuracy: 0.0000e+00 - val_loss: 14.1897 - val_accuracy: 0.0000e+00\n",
            "Epoch 21/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 13.2580 - accuracy: 0.0000e+00 - val_loss: 11.7072 - val_accuracy: 0.0000e+00\n",
            "Epoch 22/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 10.5503 - accuracy: 0.0000e+00 - val_loss: 7.8400 - val_accuracy: 0.0000e+00\n",
            "Epoch 23/100\n",
            "30/30 [==============================] - 2s 50ms/step - loss: 6.2330 - accuracy: 0.0000e+00 - val_loss: 3.1898 - val_accuracy: 0.0000e+00\n",
            "Epoch 24/100\n",
            "30/30 [==============================] - 2s 66ms/step - loss: 3.1325 - accuracy: 0.0000e+00 - val_loss: 1.9434 - val_accuracy: 0.2404\n",
            "Epoch 25/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 2.0303 - accuracy: 0.0662 - val_loss: 1.6470 - val_accuracy: 0.2788\n",
            "Epoch 26/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 1.4412 - accuracy: 0.2457 - val_loss: 1.1405 - val_accuracy: 0.6058\n",
            "Epoch 27/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.9899 - accuracy: 0.4338 - val_loss: 0.7410 - val_accuracy: 0.6154\n",
            "Epoch 28/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 0.7039 - accuracy: 0.5588 - val_loss: 0.5601 - val_accuracy: 0.6827\n",
            "Epoch 29/100\n",
            "30/30 [==============================] - 2s 63ms/step - loss: 0.4750 - accuracy: 0.7276 - val_loss: 0.2887 - val_accuracy: 0.9423\n",
            "Epoch 30/100\n",
            "30/30 [==============================] - 2s 76ms/step - loss: 0.3354 - accuracy: 0.8568 - val_loss: 0.2207 - val_accuracy: 0.9712\n",
            "Epoch 31/100\n",
            "30/30 [==============================] - 1s 45ms/step - loss: 0.2787 - accuracy: 0.8878 - val_loss: 0.2062 - val_accuracy: 0.9808\n",
            "Epoch 32/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.2022 - accuracy: 0.9316 - val_loss: 0.1308 - val_accuracy: 0.9904\n",
            "Epoch 33/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 0.1932 - accuracy: 0.9509 - val_loss: 0.1518 - val_accuracy: 0.9808\n",
            "Epoch 34/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.1389 - accuracy: 0.9658 - val_loss: 0.1160 - val_accuracy: 0.9904\n",
            "Epoch 35/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 0.1202 - accuracy: 0.9690 - val_loss: 0.1201 - val_accuracy: 0.9808\n",
            "Epoch 36/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.1079 - accuracy: 0.9754 - val_loss: 0.1157 - val_accuracy: 0.9904\n",
            "Epoch 37/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 0.1007 - accuracy: 0.9765 - val_loss: 0.1109 - val_accuracy: 0.9904\n",
            "Epoch 38/100\n",
            "30/30 [==============================] - 2s 59ms/step - loss: 0.0762 - accuracy: 0.9882 - val_loss: 0.1084 - val_accuracy: 0.9904\n",
            "Epoch 39/100\n",
            "30/30 [==============================] - 2s 65ms/step - loss: 0.0883 - accuracy: 0.9786 - val_loss: 0.1241 - val_accuracy: 0.9808\n",
            "Epoch 40/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.0916 - accuracy: 0.9765 - val_loss: 0.0971 - val_accuracy: 0.9904\n",
            "Epoch 41/100\n",
            "30/30 [==============================] - 1s 43ms/step - loss: 0.0634 - accuracy: 0.9861 - val_loss: 0.1062 - val_accuracy: 0.9904\n",
            "Epoch 42/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.0702 - accuracy: 0.9765 - val_loss: 0.1138 - val_accuracy: 0.9808\n",
            "Epoch 43/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.0586 - accuracy: 0.9861 - val_loss: 0.0978 - val_accuracy: 0.9904\n",
            "Epoch 44/100\n",
            "30/30 [==============================] - 1s 42ms/step - loss: 0.1305 - accuracy: 0.9647 - val_loss: 0.0899 - val_accuracy: 0.9904\n",
            "Epoch 45/100\n",
            "30/30 [==============================] - 1s 41ms/step - loss: 0.0916 - accuracy: 0.9744 - val_loss: 0.1008 - val_accuracy: 0.9904\n",
            "Epoch 46/100\n",
            "30/30 [==============================] - 2s 65ms/step - loss: 0.0423 - accuracy: 0.9925 - val_loss: 0.0979 - val_accuracy: 0.9904\n",
            "Epoch 47/100\n",
            "30/30 [==============================] - 2s 63ms/step - loss: 0.0766 - accuracy: 0.9840 - val_loss: 0.0957 - val_accuracy: 0.9904\n",
            "Epoch 48/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.0501 - accuracy: 0.9882 - val_loss: 0.1004 - val_accuracy: 0.9904\n",
            "Epoch 49/100\n",
            "30/30 [==============================] - 1s 44ms/step - loss: 0.0706 - accuracy: 0.9744 - val_loss: 0.1007 - val_accuracy: 0.9904\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 0 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# model 테스트 해 보기.\n",
        "\n",
        "test_data_dir = Path(\"/content/drive/MyDrive/input/samples\")\n",
        "\n",
        "# Get list of all the images\n",
        "images = list(test_data_dir.glob(\"*.jpg\"))\n",
        "print(\"Number of images found: \", len(images))\n",
        "\n",
        "  # Store all the characters in a set\n",
        "characters = set()\n",
        "\n",
        "  # A list to store the length of each captcha\n",
        "captcha_length = []\n",
        "\n",
        "  # Store image-label info\n",
        "dataset = []\n",
        "\n",
        "  # Iterate over the dataset and store the\n",
        "  # information needed\n",
        "for img_path in images:\n",
        "  # 1. Get the label associated with each image\n",
        "  if str(img_path)[-3:] == 'png':\n",
        "    label = img_path.name.split(\".png\")[0]\n",
        "  else:\n",
        "    label = img_path.name.split(\".jpg\")[0]\n",
        "  # 2. Store the length of this cpatcha\n",
        "  captcha_length.append(len(label))\n",
        "  # 3. Store the image-label pair info\n",
        "  dataset.append((str(img_path), label))    \n",
        "  # 4. Store the characters present\n",
        "  for ch in label:\n",
        "    characters.add(ch)\n",
        "\n",
        "  # Sort the characters        \n",
        "characters = sorted(characters)\n",
        "\n",
        "  # Convert the dataset info into a dataframe\n",
        "dataset = pd.DataFrame(dataset, columns=[\"img_path\", \"label\"], index=None)\n",
        "\n",
        "  # Shuffle the dataset\n",
        "dataset = dataset.sample(frac=1.).reset_index(drop=True)\n",
        "\n",
        "  # Build test data\n",
        "test_data, test_labels = generate_arrays(df=dataset)\n",
        "print(\"Number of test images: \", test_data.shape)\n",
        "print(\"Number of test labels: \", test_labels.shape)\n",
        "\n",
        "  # Map text to numeric labels \n",
        "char_to_labels = {char:idx for idx, char in enumerate(characters)}\n",
        "\n",
        "  # Map numeric labels to text\n",
        "labels_to_char = {val:key for key, val in char_to_labels.items()}\n",
        "\n",
        "  # Batch size for training and validation\n",
        "batch_size = 32\n",
        "\n",
        "  # Desired image dimensions\n",
        "img_width=200\n",
        "img_height=50 \n",
        "\n",
        "  # Factor  by which the image is going to be downsampled\n",
        "  # by the convolutional blocks\n",
        "downsample_factor=4\n",
        "\n",
        "  # Maximum length of any captcha in the data\n",
        "  # 글자 개수에 맞춰서 4개 or 5개\n",
        "max_length=5\n",
        "\n",
        "  # Get a generator object for the validation data \n",
        "test_data_generator = DataGenerator(data=test_data,\n",
        "                                      labels=test_labels,\n",
        "                                      char_map=char_to_labels,\n",
        "                                      batch_size=batch_size,\n",
        "                                      img_width=img_width,\n",
        "                                      img_height=img_height,\n",
        "                                      downsample_factor=downsample_factor,\n",
        "                                      max_length=max_length,\n",
        "                                      shuffle=False\n",
        "                                      )\n",
        "\n",
        "model_addr = '/content/drive/MyDrive/captcha_recognition_model.h5' # 모델 주소 입력\n",
        "test_model = tf.keras.models.load_model(model_addr, custom_objects={\"CTCLayer\": CTCLayer})\n",
        "print(test_model.evaluate(test_data_generator))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "R3CvMd8onSAd",
        "outputId": "868ca1d6-70a2-497d-ec3e-86db9f92c56c"
      },
      "execution_count": 96,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of images found:  30\n",
            "Number of test images:  (30, 50, 200)\n",
            "Number of test labels:  (30,)\n",
            "1/1 [==============================] - 2s 2s/step - loss: 0.0428 - accuracy: 1.0000\n",
            "[0.04283486679196358, 1.0]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "prediction_model = keras.models.Model(test_model.get_layer(name='input_data').input,\n",
        "                                        test_model.get_layer(name='dense2').output)\n",
        "\n",
        "# A utility to decode the output of the network\n",
        "def decode_batch_predictions(pred):\n",
        "    pred = pred[:, :-2]\n",
        "    input_len = np.ones(pred.shape[0])*pred.shape[1]\n",
        "    \n",
        "    # Use greedy search. For complex tasks, you can use beam search\n",
        "    results = keras.backend.ctc_decode(pred, \n",
        "                                        input_length=input_len,\n",
        "                                        greedy=True)[0][0]\n",
        "    \n",
        "    # Iterate over the results and get back the text\n",
        "    output_text = []\n",
        "    for res in results.numpy():\n",
        "        outstr = ''\n",
        "        for c in res:\n",
        "            if c < len(characters) and c >=0:\n",
        "                outstr += labels_to_char[c]\n",
        "        output_text.append(outstr)\n",
        "    \n",
        "    # return final text results\n",
        "    return output_text\n",
        "\n",
        "p_texts = []\n",
        "o_texts = []\n",
        "\n",
        "#  Let's check results on some validation samples\n",
        "for p, (inp_value, _) in enumerate(test_data_generator):\n",
        "    bs = inp_value['input_data'].shape[0]\n",
        "    X_data = inp_value['input_data']\n",
        "    labels = inp_value['input_label']\n",
        "    \n",
        "    preds = prediction_model.predict(X_data)\n",
        "    pred_texts = decode_batch_predictions(preds)\n",
        "    \n",
        "    \n",
        "    orig_texts = []\n",
        "    for label in labels:\n",
        "        text = ''.join([labels_to_char[int(x)] for x in label])\n",
        "        orig_texts.append(text)\n",
        "\n",
        "    p_texts.append(pred_texts)\n",
        "    o_texts.append(orig_texts)\n",
        "    \n",
        "    for i in range(bs):\n",
        "        print(f'Ground truth: {orig_texts[i]} \\t Predicted: {pred_texts[i]} \\t Correct: {orig_texts[i]==pred_texts[i]}')\n",
        "    break"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "baavdBCyTIsy",
        "outputId": "641c8384-d5cb-4549-9452-3a598efdcdbe"
      },
      "execution_count": 98,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 33s 33s/step\n",
            "Ground truth: 85pew \t Predicted: 85pew \t Correct: True\n",
            "Ground truth: ygce8 \t Predicted: ygce8 \t Correct: True\n",
            "Ground truth: gmmne \t Predicted: gmmne \t Correct: True\n",
            "Ground truth: n3xfg \t Predicted: n3xfg \t Correct: True\n",
            "Ground truth: fgb36 \t Predicted: fgb36 \t Correct: True\n",
            "Ground truth: bgxcd \t Predicted: bgxcd \t Correct: True\n",
            "Ground truth: w6ny4 \t Predicted: w6ny4 \t Correct: True\n",
            "Ground truth: yge7c \t Predicted: yge7c \t Correct: True\n",
            "Ground truth: ewyg7 \t Predicted: ewyg7 \t Correct: True\n",
            "Ground truth: f6bpw \t Predicted: f6bpw \t Correct: True\n",
            "Ground truth: pcmcc \t Predicted: pcmcc \t Correct: True\n",
            "Ground truth: 62fgn \t Predicted: 62fgn \t Correct: True\n",
            "Ground truth: nxf2c \t Predicted: nxf2c \t Correct: True\n",
            "Ground truth: e3cfe \t Predicted: e3cfe \t Correct: True\n",
            "Ground truth: 2nbc5 \t Predicted: 2nbc5 \t Correct: True\n",
            "Ground truth: b4ndb \t Predicted: b4ndb \t Correct: True\n",
            "Ground truth: yw8y5 \t Predicted: yw8y5 \t Correct: True\n",
            "Ground truth: n8wxm \t Predicted: n8wxm \t Correct: True\n",
            "Ground truth: geyn5 \t Predicted: geyn5 \t Correct: True\n",
            "Ground truth: 78dw6 \t Predicted: 78dw6 \t Correct: True\n",
            "Ground truth: cndmc \t Predicted: cndmc \t Correct: True\n",
            "Ground truth: ddxpp \t Predicted: ddxpp \t Correct: True\n",
            "Ground truth: d2n8x \t Predicted: d2n8x \t Correct: True\n",
            "Ground truth: nn6w6 \t Predicted: nn6w6 \t Correct: True\n",
            "Ground truth: bx5ed \t Predicted: bx5ed \t Correct: True\n",
            "Ground truth: dgcm4 \t Predicted: dgcm4 \t Correct: True\n",
            "Ground truth: n2c85 \t Predicted: n2c85 \t Correct: True\n",
            "Ground truth: 65nmw \t Predicted: 65nmw \t Correct: True\n",
            "Ground truth: 84py4 \t Predicted: 84py4 \t Correct: True\n",
            "Ground truth: 3c7de \t Predicted: 3c7de \t Correct: True\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#  Let's check results on some validation samples\n",
        "for p, (inp_value, _) in enumerate(valid_data_generator):\n",
        "    bs = inp_value['input_data'].shape[0]\n",
        "    X_data = inp_value['input_data']\n",
        "    labels = inp_value['input_label']\n",
        "    \n",
        "    preds = prediction_model.predict(X_data)\n",
        "    pred_texts = decode_batch_predictions(preds)\n",
        "    \n",
        "    \n",
        "    orig_texts = []\n",
        "    for label in labels:\n",
        "        text = ''.join([labels_to_char[int(x)] for x in label])\n",
        "        orig_texts.append(text)\n",
        "        \n",
        "    for i in range(bs):\n",
        "        print(f'Ground truth: {orig_texts[i]} \\t Predicted: {pred_texts[i]}')\n",
        "    break"
      ],
      "metadata": {
        "id": "Plj8LfMbS1b-"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Please upvote if you liked the notebook.**"
      ],
      "metadata": {
        "id": "mmcxN9vL4A5C"
      }
    }
  ]
}