Sapana_Dashoni_Google_Hacathon_SWE_2025.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "toc_visible": true,
      "authorship_tag": "ABX9TyMxCDxL7aY0ZY+Tc9VoTRqb",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/SapanaDashoni15/Algo-Visualizer/blob/main/Sapana_Dashoni_Google_Hacathon_SWE_2025.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Mount Google Drive"
      ],
      "metadata": {
        "id": "BCYOcjdrtv0R"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZFzGoZMntOrs",
        "outputId": "17fbd36e-bec3-44b9-9777-0d83381319f9"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Set Dataset Path"
      ],
      "metadata": {
        "id": "Curm_qoqt2Oh"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import os\n",
        "\n",
        "dataset_path = \"/content/drive/MyDrive/YoloDataset\"  # Adjust this path\n",
        "train_path_images = os.path.join(dataset_path, \"train/images\")\n",
        "train_path_labels = os.path.join(dataset_path, \"train/labels\")\n",
        "valid_path_images = os.path.join(dataset_path, \"valid/images\")\n",
        "valid_path_labels = os.path.join(dataset_path, \"valid/labels\")\n",
        "test_path_images = os.path.join(dataset_path, \"test/images\")\n",
        "test_path_labels = os.path.join(dataset_path, \"test/labels\")\n"
      ],
      "metadata": {
        "id": "VMHbX2lNt6Nq"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Check Dataset Files"
      ],
      "metadata": {
        "id": "9Z5n7Be6uCvZ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import glob\n",
        "\n",
        "train_images = sorted(glob.glob(os.path.join(train_path_images, \"*.jpg\")))\n",
        "train_labels = sorted(glob.glob(os.path.join(train_path_labels, \"*.txt\")))\n",
        "\n",
        "valid_images = sorted(glob.glob(os.path.join(valid_path_images, \"*.jpg\")))\n",
        "valid_labels = sorted(glob.glob(os.path.join(valid_path_labels, \"*.txt\")))\n",
        "\n",
        "test_images = sorted(glob.glob(os.path.join(test_path_images, \"*.jpg\")))\n",
        "test_labels = sorted(glob.glob(os.path.join(test_path_labels, \"*.txt\")))\n",
        "\n",
        "print(f\"Train Images: {len(train_images)}, Train Labels: {len(train_labels)}\")\n",
        "print(f\"Valid Images: {len(valid_images)}, Valid Labels: {len(valid_labels)}\")\n",
        "print(f\"Test Images: {len(test_images)}, Test Labels: {len(test_labels)}\")\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3HDsRf80t-qi",
        "outputId": "29a3c812-433d-45ec-eb0f-4ca4f6bf4f6e"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Train Images: 374, Train Labels: 374\n",
            "Valid Images: 75, Valid Labels: 75\n",
            "Test Images: 40, Test Labels: 40\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Model Development (CNN + BiLSTM + CTC Loss)"
      ],
      "metadata": {
        "id": "N19r4NApuGQ5"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install tensorflow keras-ocr jiwer\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Du5z00ONuJaP",
        "outputId": "3cd9f2e8-f606-4d7e-85fb-ae1565d9389d"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: tensorflow in /usr/local/lib/python3.11/dist-packages (2.18.0)\n",
            "Collecting keras-ocr\n",
            "  Downloading keras_ocr-0.9.3-py3-none-any.whl.metadata (8.6 kB)\n",
            "Collecting jiwer\n",
            "  Downloading jiwer-3.1.0-py3-none-any.whl.metadata (2.6 kB)\n",
            "Requirement already satisfied: absl-py>=1.0.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (1.4.0)\n",
            "Requirement already satisfied: astunparse>=1.6.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (1.6.3)\n",
            "Requirement already satisfied: flatbuffers>=24.3.25 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (25.2.10)\n",
            "Requirement already satisfied: gast!=0.5.0,!=0.5.1,!=0.5.2,>=0.2.1 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (0.6.0)\n",
            "Requirement already satisfied: google-pasta>=0.1.1 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (0.2.0)\n",
            "Requirement already satisfied: libclang>=13.0.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (18.1.1)\n",
            "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (3.4.0)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.11/dist-packages (from tensorflow) (24.2)\n",
            "Requirement already satisfied: protobuf!=4.21.0,!=4.21.1,!=4.21.2,!=4.21.3,!=4.21.4,!=4.21.5,<6.0.0dev,>=3.20.3 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (4.25.6)\n",
            "Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (2.32.3)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.11/dist-packages (from tensorflow) (75.1.0)\n",
            "Requirement already satisfied: six>=1.12.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (1.17.0)\n",
            "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (2.5.0)\n",
            "Requirement already satisfied: typing-extensions>=3.6.6 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (4.12.2)\n",
            "Requirement already satisfied: wrapt>=1.11.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (1.17.2)\n",
            "Requirement already satisfied: grpcio<2.0,>=1.24.3 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (1.70.0)\n",
            "Requirement already satisfied: tensorboard<2.19,>=2.18 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (2.18.0)\n",
            "Requirement already satisfied: keras>=3.5.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (3.8.0)\n",
            "Requirement already satisfied: numpy<2.1.0,>=1.26.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (1.26.4)\n",
            "Requirement already satisfied: h5py>=3.11.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (3.12.1)\n",
            "Requirement already satisfied: ml-dtypes<0.5.0,>=0.4.0 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (0.4.1)\n",
            "Requirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in /usr/local/lib/python3.11/dist-packages (from tensorflow) (0.37.1)\n",
            "Requirement already satisfied: editdistance in /usr/local/lib/python3.11/dist-packages (from keras-ocr) (0.8.1)\n",
            "Collecting efficientnet==1.0.0 (from keras-ocr)\n",
            "  Downloading efficientnet-1.0.0-py3-none-any.whl.metadata (6.1 kB)\n",
            "Collecting essential_generators (from keras-ocr)\n",
            "  Downloading essential_generators-1.0-py3-none-any.whl.metadata (14 kB)\n",
            "Requirement already satisfied: fonttools in /usr/local/lib/python3.11/dist-packages (from keras-ocr) (4.56.0)\n",
            "Requirement already satisfied: imgaug in /usr/local/lib/python3.11/dist-packages (from keras-ocr) (0.4.0)\n",
            "Collecting pyclipper (from keras-ocr)\n",
            "  Downloading pyclipper-1.3.0.post6-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (9.0 kB)\n",
            "Requirement already satisfied: shapely in /usr/local/lib/python3.11/dist-packages (from keras-ocr) (2.0.7)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.11/dist-packages (from keras-ocr) (4.67.1)\n",
            "Collecting validators (from keras-ocr)\n",
            "  Downloading validators-0.34.0-py3-none-any.whl.metadata (3.8 kB)\n",
            "Collecting keras-applications<=1.0.8,>=1.0.7 (from efficientnet==1.0.0->keras-ocr)\n",
            "  Downloading Keras_Applications-1.0.8-py3-none-any.whl.metadata (1.7 kB)\n",
            "Requirement already satisfied: scikit-image in /usr/local/lib/python3.11/dist-packages (from efficientnet==1.0.0->keras-ocr) (0.25.2)\n",
            "Requirement already satisfied: click>=8.1.8 in /usr/local/lib/python3.11/dist-packages (from jiwer) (8.1.8)\n",
            "Collecting rapidfuzz>=3.9.7 (from jiwer)\n",
            "  Downloading rapidfuzz-3.12.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (11 kB)\n",
            "Requirement already satisfied: wheel<1.0,>=0.23.0 in /usr/local/lib/python3.11/dist-packages (from astunparse>=1.6.0->tensorflow) (0.45.1)\n",
            "Requirement already satisfied: rich in /usr/local/lib/python3.11/dist-packages (from keras>=3.5.0->tensorflow) (13.9.4)\n",
            "Requirement already satisfied: namex in /usr/local/lib/python3.11/dist-packages (from keras>=3.5.0->tensorflow) (0.0.8)\n",
            "Requirement already satisfied: optree in /usr/local/lib/python3.11/dist-packages (from keras>=3.5.0->tensorflow) (0.14.0)\n",
            "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.11/dist-packages (from requests<3,>=2.21.0->tensorflow) (3.4.1)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.11/dist-packages (from requests<3,>=2.21.0->tensorflow) (3.10)\n",
            "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.11/dist-packages (from requests<3,>=2.21.0->tensorflow) (2.3.0)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.11/dist-packages (from requests<3,>=2.21.0->tensorflow) (2025.1.31)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.11/dist-packages (from tensorboard<2.19,>=2.18->tensorflow) (3.7)\n",
            "Requirement already satisfied: tensorboard-data-server<0.8.0,>=0.7.0 in /usr/local/lib/python3.11/dist-packages (from tensorboard<2.19,>=2.18->tensorflow) (0.7.2)\n",
            "Requirement already satisfied: werkzeug>=1.0.1 in /usr/local/lib/python3.11/dist-packages (from tensorboard<2.19,>=2.18->tensorflow) (3.1.3)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.11/dist-packages (from imgaug->keras-ocr) (1.13.1)\n",
            "Requirement already satisfied: Pillow in /usr/local/lib/python3.11/dist-packages (from imgaug->keras-ocr) (11.1.0)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.11/dist-packages (from imgaug->keras-ocr) (3.10.0)\n",
            "Requirement already satisfied: opencv-python in /usr/local/lib/python3.11/dist-packages (from imgaug->keras-ocr) (4.11.0.86)\n",
            "Requirement already satisfied: imageio in /usr/local/lib/python3.11/dist-packages (from imgaug->keras-ocr) (2.37.0)\n",
            "Requirement already satisfied: networkx>=3.0 in /usr/local/lib/python3.11/dist-packages (from scikit-image->efficientnet==1.0.0->keras-ocr) (3.4.2)\n",
            "Requirement already satisfied: tifffile>=2022.8.12 in /usr/local/lib/python3.11/dist-packages (from scikit-image->efficientnet==1.0.0->keras-ocr) (2025.2.18)\n",
            "Requirement already satisfied: lazy-loader>=0.4 in /usr/local/lib/python3.11/dist-packages (from scikit-image->efficientnet==1.0.0->keras-ocr) (0.4)\n",
            "Requirement already satisfied: MarkupSafe>=2.1.1 in /usr/local/lib/python3.11/dist-packages (from werkzeug>=1.0.1->tensorboard<2.19,>=2.18->tensorflow) (3.0.2)\n",
            "Requirement already satisfied: contourpy>=1.0.1 in /usr/local/lib/python3.11/dist-packages (from matplotlib->imgaug->keras-ocr) (1.3.1)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.11/dist-packages (from matplotlib->imgaug->keras-ocr) (0.12.1)\n",
            "Requirement already satisfied: kiwisolver>=1.3.1 in /usr/local/lib/python3.11/dist-packages (from matplotlib->imgaug->keras-ocr) (1.4.8)\n",
            "Requirement already satisfied: pyparsing>=2.3.1 in /usr/local/lib/python3.11/dist-packages (from matplotlib->imgaug->keras-ocr) (3.2.1)\n",
            "Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.11/dist-packages (from matplotlib->imgaug->keras-ocr) (2.8.2)\n",
            "Requirement already satisfied: markdown-it-py>=2.2.0 in /usr/local/lib/python3.11/dist-packages (from rich->keras>=3.5.0->tensorflow) (3.0.0)\n",
            "Requirement already satisfied: pygments<3.0.0,>=2.13.0 in /usr/local/lib/python3.11/dist-packages (from rich->keras>=3.5.0->tensorflow) (2.18.0)\n",
            "Requirement already satisfied: mdurl~=0.1 in /usr/local/lib/python3.11/dist-packages (from markdown-it-py>=2.2.0->rich->keras>=3.5.0->tensorflow) (0.1.2)\n",
            "Downloading keras_ocr-0.9.3-py3-none-any.whl (42 kB)\n",
            "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m42.3/42.3 kB\u001b[0m \u001b[31m4.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hDownloading efficientnet-1.0.0-py3-none-any.whl (17 kB)\n",
            "Downloading jiwer-3.1.0-py3-none-any.whl (22 kB)\n",
            "Downloading rapidfuzz-3.12.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (3.1 MB)\n",
            "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m3.1/3.1 MB\u001b[0m \u001b[31m56.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hDownloading essential_generators-1.0-py3-none-any.whl (9.5 MB)\n",
            "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m9.5/9.5 MB\u001b[0m \u001b[31m60.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hDownloading pyclipper-1.3.0.post6-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (969 kB)\n",
            "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m969.6/969.6 kB\u001b[0m \u001b[31m51.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hDownloading validators-0.34.0-py3-none-any.whl (43 kB)\n",
            "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m43.5/43.5 kB\u001b[0m \u001b[31m4.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hDownloading Keras_Applications-1.0.8-py3-none-any.whl (50 kB)\n",
            "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m50.7/50.7 kB\u001b[0m \u001b[31m4.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hInstalling collected packages: pyclipper, essential_generators, validators, rapidfuzz, keras-applications, jiwer, efficientnet, keras-ocr\n",
            "Successfully installed efficientnet-1.0.0 essential_generators-1.0 jiwer-3.1.0 keras-applications-1.0.8 keras-ocr-0.9.3 pyclipper-1.3.0.post6 rapidfuzz-3.12.1 validators-0.34.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import cv2\n",
        "import numpy as np\n",
        "\n",
        "def preprocess_image(image_path):\n",
        "    img = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)  # Load as grayscale\n",
        "    img = cv2.resize(img, (128, 32))  # Resize to match model input\n",
        "    img = img / 255.0  # Normalize pixel values\n",
        "    img = np.expand_dims(img, axis=-1)  # Add channel dimension (H, W, 1)\n",
        "    return img\n"
      ],
      "metadata": {
        "id": "5kBtO7QQuMMK"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras.preprocessing.text import Tokenizer\n",
        "\n",
        "# Tokenize characters\n",
        "char_tokenizer = Tokenizer(char_level=True, filters='')\n",
        "\n",
        "# Collect all labels\n",
        "all_texts = [\" \".join(open(label).read().strip()) for label in train_labels]\n",
        "char_tokenizer.fit_on_texts(all_texts)\n",
        "\n",
        "# Convert text to sequences\n",
        "def text_to_sequence(text):\n",
        "    return char_tokenizer.texts_to_sequences([text])[0]\n"
      ],
      "metadata": {
        "id": "PrWuKZJluOuJ"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Define CNN + BiLSTM Model"
      ],
      "metadata": {
        "id": "v2xDAySH7bRr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow.keras import layers, models\n",
        "\n",
        "def build_model(input_shape, num_classes):\n",
        "    inputs = layers.Input(shape=input_shape)\n",
        "\n",
        "    # CNN Feature Extractor\n",
        "    x = layers.Conv2D(32, (3,3), activation='relu', padding='same')(inputs)\n",
        "    x = layers.MaxPooling2D(pool_size=(2,2))(x)\n",
        "    x = layers.Conv2D(64, (3,3), activation='relu', padding='same')(x)\n",
        "    x = layers.MaxPooling2D(pool_size=(2,2))(x)\n",
        "\n",
        "    # Reshape for LSTM\n",
        "    x = layers.Reshape(target_shape=(-1, x.shape[-1]))(x)\n",
        "\n",
        "    # BiLSTM Layers\n",
        "    x = layers.Bidirectional(layers.LSTM(128, return_sequences=True))(x)\n",
        "    x = layers.Bidirectional(layers.LSTM(128, return_sequences=True))(x)\n",
        "\n",
        "    # Output Layer (Character Prediction)\n",
        "    x = layers.Dense(num_classes+1, activation='softmax')(x)\n",
        "\n",
        "    model = models.Model(inputs, x)\n",
        "    return model\n",
        "\n",
        "# Get number of characters in dataset\n",
        "num_chars = len(char_tokenizer.word_index)\n",
        "input_shape = (128, 32, 1)\n",
        "model = build_model(input_shape, num_chars)\n",
        "model.summary()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 434
        },
        "id": "VjX-8A2buWe-",
        "outputId": "d661e68d-260f-4582-e8dc-90eaa931373e"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1mModel: \"functional\"\u001b[0m\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"functional\"</span>\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                        \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape               \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m        Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ input_layer (\u001b[38;5;33mInputLayer\u001b[0m)             │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m1\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d (\u001b[38;5;33mConv2D\u001b[0m)                      │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m)         │             \u001b[38;5;34m320\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d (\u001b[38;5;33mMaxPooling2D\u001b[0m)         │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m32\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_1 (\u001b[38;5;33mConv2D\u001b[0m)                    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m64\u001b[0m)          │          \u001b[38;5;34m18,496\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_1 (\u001b[38;5;33mMaxPooling2D\u001b[0m)       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m64\u001b[0m)           │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ reshape (\u001b[38;5;33mReshape\u001b[0m)                    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m64\u001b[0m)             │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional (\u001b[38;5;33mBidirectional\u001b[0m)        │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m256\u001b[0m)            │         \u001b[38;5;34m197,632\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional_1 (\u001b[38;5;33mBidirectional\u001b[0m)      │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m256\u001b[0m)            │         \u001b[38;5;34m394,240\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense (\u001b[38;5;33mDense\u001b[0m)                        │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m14\u001b[0m)             │           \u001b[38;5;34m3,598\u001b[0m │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃<span style=\"font-weight: bold\"> Layer (type)                         </span>┃<span style=\"font-weight: bold\"> Output Shape                </span>┃<span style=\"font-weight: bold\">         Param # </span>┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ input_layer (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)             │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv2D</span>)                      │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)         │             <span style=\"color: #00af00; text-decoration-color: #00af00\">320</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">MaxPooling2D</span>)         │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">16</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv2D</span>)                    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">16</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)          │          <span style=\"color: #00af00; text-decoration-color: #00af00\">18,496</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">MaxPooling2D</span>)       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">8</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)           │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ reshape (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Reshape</span>)                    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)             │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Bidirectional</span>)        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>)            │         <span style=\"color: #00af00; text-decoration-color: #00af00\">197,632</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Bidirectional</span>)      │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>)            │         <span style=\"color: #00af00; text-decoration-color: #00af00\">394,240</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                        │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">14</span>)             │           <span style=\"color: #00af00; text-decoration-color: #00af00\">3,598</span> │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m614,286\u001b[0m (2.34 MB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">614,286</span> (2.34 MB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m614,286\u001b[0m (2.34 MB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">614,286</span> (2.34 MB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Build the CTC Model"
      ],
      "metadata": {
        "id": "BYnbls4aC62L"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow.keras import layers, models, Input\n",
        "from tensorflow.keras import backend as K\n",
        "from tensorflow.keras.layers import Lambda\n",
        "\n",
        "# 1. Build the Base Model (CNN + BiLSTM)\n",
        "input_img = Input(shape=(128, 32, 1), name='input_img')  # Expected input shape\n",
        "\n",
        "# CNN part\n",
        "x = layers.Conv2D(32, (3,3), activation='relu', padding='same')(input_img)\n",
        "x = layers.MaxPooling2D(pool_size=(2,2))(x)\n",
        "x = layers.Conv2D(64, (3,3), activation='relu', padding='same')(x)\n",
        "x = layers.MaxPooling2D(pool_size=(2,2))(x)\n",
        "\n",
        "# At this point, x might have shape (batch, H, W, channels).\n",
        "# We need to reshape it to (batch, time_steps, features) for the LSTM.\n",
        "# For example, assume we want a fixed time_steps of 256.\n",
        "# We'll use a Lambda layer to resize the feature map's height to 256.\n",
        "x = layers.Lambda(lambda x: tf.image.resize(x, (256, tf.shape(x)[2])))(x)\n",
        "\n",
        "# Reshape to (batch, time_steps, features)\n",
        "x = layers.Reshape(target_shape=(256, -1))(x)\n",
        "\n",
        "# Add BiLSTM layers\n",
        "x = layers.Bidirectional(layers.LSTM(128, return_sequences=True))(x)\n",
        "x = layers.Bidirectional(layers.LSTM(128, return_sequences=True))(x)\n",
        "\n",
        "# Final Dense layer to produce predictions over classes\n",
        "# Suppose we have 14 classes; CTC requires one extra (blank) class\n",
        "num_classes = 14\n",
        "y_pred = layers.Dense(num_classes + 1, activation='softmax', name='y_pred')(x)\n",
        "\n",
        "# Define the base model (for reference)\n",
        "base_model = models.Model(inputs=input_img, outputs=y_pred)\n",
        "print(\"Base Model Summary:\")\n",
        "base_model.summary()\n",
        "\n",
        "# 2. Build the CTC Model\n",
        "\n",
        "# Additional inputs required for CTC loss:\n",
        "labels = Input(name='labels', shape=[None], dtype='int32')       # Ground truth sequences (variable length)\n",
        "input_length = Input(name='input_length', shape=[1], dtype='int32')  # Length of the predicted sequence (e.g., 256)\n",
        "label_length = Input(name='label_length', shape=[1], dtype='int32')  # Actual lengths of the ground truth sequences\n",
        "\n",
        "# CTC Loss Lambda function\n",
        "def ctc_lambda_func(args):\n",
        "    y_pred, labels, input_length, label_length = args\n",
        "    # y_pred should be in log probabilities; using softmax activation means it's fine for K.ctc_batch_cost\n",
        "    return K.ctc_batch_cost(labels, y_pred, input_length, label_length)\n",
        "\n",
        "# Create the Lambda layer for CTC loss computation; this produces the loss output\n",
        "loss_out = Lambda(ctc_lambda_func, output_shape=(1,), name='ctc')([y_pred, labels, input_length, label_length])\n",
        "\n",
        "# Define the complete model for training with CTC loss\n",
        "ctc_model = models.Model(inputs=[input_img, labels, input_length, label_length], outputs=loss_out)\n",
        "\n",
        "# 3. Compile the CTC Model\n",
        "# We use a dummy lambda loss since the output of our model is the computed loss.\n",
        "ctc_model.compile(optimizer='adam', loss={'ctc': lambda y_true, y_pred: y_pred})\n",
        "\n",
        "print(\"CTC Model is compiled and ready for training.\")\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 501
        },
        "id": "Dtdy0W60DNGy",
        "outputId": "be265764-b6cc-4ea9-9a22-b525aecdabe9"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Base Model Summary:\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1mModel: \"functional_1\"\u001b[0m\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"functional_1\"</span>\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                        \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape               \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m        Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ input_img (\u001b[38;5;33mInputLayer\u001b[0m)               │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m1\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_2 (\u001b[38;5;33mConv2D\u001b[0m)                    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m)         │             \u001b[38;5;34m320\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_2 (\u001b[38;5;33mMaxPooling2D\u001b[0m)       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m32\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_3 (\u001b[38;5;33mConv2D\u001b[0m)                    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m64\u001b[0m)          │          \u001b[38;5;34m18,496\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_3 (\u001b[38;5;33mMaxPooling2D\u001b[0m)       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m64\u001b[0m)           │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ lambda (\u001b[38;5;33mLambda\u001b[0m)                      │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m64\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ reshape_1 (\u001b[38;5;33mReshape\u001b[0m)                  │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m512\u001b[0m)            │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional_2 (\u001b[38;5;33mBidirectional\u001b[0m)      │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m256\u001b[0m)            │         \u001b[38;5;34m656,384\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional_3 (\u001b[38;5;33mBidirectional\u001b[0m)      │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m256\u001b[0m)            │         \u001b[38;5;34m394,240\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ y_pred (\u001b[38;5;33mDense\u001b[0m)                       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m256\u001b[0m, \u001b[38;5;34m15\u001b[0m)             │           \u001b[38;5;34m3,855\u001b[0m │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃<span style=\"font-weight: bold\"> Layer (type)                         </span>┃<span style=\"font-weight: bold\"> Output Shape                </span>┃<span style=\"font-weight: bold\">         Param # </span>┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ input_img (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_2 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv2D</span>)                    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)         │             <span style=\"color: #00af00; text-decoration-color: #00af00\">320</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_2 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">MaxPooling2D</span>)       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">16</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_3 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv2D</span>)                    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">16</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)          │          <span style=\"color: #00af00; text-decoration-color: #00af00\">18,496</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_3 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">MaxPooling2D</span>)       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">8</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)           │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ lambda (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Lambda</span>)                      │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">8</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ reshape_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Reshape</span>)                  │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">512</span>)            │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional_2 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Bidirectional</span>)      │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>)            │         <span style=\"color: #00af00; text-decoration-color: #00af00\">656,384</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ bidirectional_3 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Bidirectional</span>)      │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>)            │         <span style=\"color: #00af00; text-decoration-color: #00af00\">394,240</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ y_pred (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">256</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">15</span>)             │           <span style=\"color: #00af00; text-decoration-color: #00af00\">3,855</span> │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m1,073,295\u001b[0m (4.09 MB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">1,073,295</span> (4.09 MB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m1,073,295\u001b[0m (4.09 MB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">1,073,295</span> (4.09 MB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CTC Model is compiled and ready for training.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "ctc_model.compile(optimizer='adam', loss={'ctc': lambda y_true, y_pred: y_pred})\n"
      ],
      "metadata": {
        "id": "Td5Ff341D06b"
      },
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Model Training"
      ],
      "metadata": {
        "id": "FvAPM9cvuaqP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "from glob import glob\n",
        "\n",
        "# Define paths for training and validation label files\n",
        "train_label_paths = sorted(glob(\"/content/drive/MyDrive/YoloDataset/train/labels/*.txt\"))\n",
        "valid_label_paths = sorted(glob(\"/content/drive/MyDrive/YoloDataset/valid/labels/*.txt\"))\n",
        "\n",
        "def load_raw_labels(label_paths):\n",
        "    \"\"\"\n",
        "    Load raw labels from a list of text file paths.\n",
        "    Each file is expected to contain a sequence of numbers (as strings),\n",
        "    possibly with floating-point values. The function converts each token\n",
        "    to a float and then to an integer.\n",
        "    \"\"\"\n",
        "    raw_labels = []\n",
        "    for path in label_paths:\n",
        "        with open(path, \"r\") as f:\n",
        "            line = f.readline().strip()\n",
        "            # Convert each token to an integer via float conversion (handles tokens like \"0.8203125\")\n",
        "            seq = [int(float(x)) for x in line.split()]\n",
        "            raw_labels.append(seq)\n",
        "    return raw_labels\n",
        "\n",
        "# Load the raw label sequences for training and validation\n",
        "train_raw_labels = load_raw_labels(train_label_paths)\n",
        "valid_raw_labels = load_raw_labels(valid_label_paths)\n",
        "\n",
        "# Check the first few raw labels for inspection\n",
        "print(\"First 5 train raw labels:\", train_raw_labels[:5])\n",
        "print(\"First 5 valid raw labels:\", valid_raw_labels[:5])\n",
        "\n",
        "# Compute the length of each label sequence (i.e. number of tokens in each label)\n",
        "train_label_length = np.array([len(seq) for seq in train_raw_labels])\n",
        "valid_label_length = np.array([len(seq) for seq in valid_raw_labels])\n",
        "\n",
        "# Assume input_length is constant (e.g., the number of timesteps output by your model, here 256)\n",
        "train_input_length = np.full((len(train_raw_labels),), 256, dtype=np.int32) # Modified to match train_raw_labels\n",
        "valid_input_length = np.full((len(valid_raw_labels),), 256, dtype=np.int32) # Modified to match valid_raw_labels\n",
        "\n",
        "# Print computed label lengths and input lengths for verification\n",
        "print(\"Train label lengths:\", train_label_length)\n",
        "print(\"Valid label lengths:\", valid_label_length)\n",
        "print(\"First 5 Train input lengths:\", train_input_length[:5])\n",
        "print(\"First 5 Valid input lengths:\", valid_input_length[:5])"
      ],
      "metadata": {
        "id": "71uMh-6qD6hc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2532154c-dbc1-43c4-8b34-9b114f47f078"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "First 5 train raw labels: [[50, 0, 0, 0, 0, 0, 0, 0, 0], [318, 0, 0, 0, 0, 0, 0, 0, 0], [318, 0, 0, 0, 0, 0, 0, 0, 0], [764, 0, 0, 0, 0, 0, 0, 0, 0], [324, 0, 0, 0, 0, 0, 0, 0, 0]]\n",
            "First 5 valid raw labels: [[628, 0, 0, 0, 0, 0, 0, 0, 0], [461, 0, 0, 0, 0, 0, 0, 0, 0], [667, 0, 0, 0, 0, 0, 0, 0, 0], [216, 0, 0, 0, 0, 0, 0, 0, 0], [216, 0, 0, 0, 0, 0, 0, 0, 0]]\n",
            "Train label lengths: [9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9]\n",
            "Valid label lengths: [9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9]\n",
            "First 5 Train input lengths: [256 256 256 256 256]\n",
            "First 5 Valid input lengths: [256 256 256 256 256]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_input_length = np.full((len(train_raw_labels), 1), 256, dtype=np.int32)\n",
        "valid_input_length = np.full((len(valid_raw_labels), 1), 256, dtype=np.int32)\n"
      ],
      "metadata": {
        "id": "vOYQY0GPD95b"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "from glob import glob\n",
        "\n",
        "# Define paths for training and validation label files\n",
        "train_label_paths = sorted(glob(\"/content/drive/MyDrive/YoloDataset/train/labels/*.txt\"))\n",
        "valid_label_paths = sorted(glob(\"/content/drive/MyDrive/YoloDataset/valid/labels/*.txt\"))\n",
        "\n",
        "def load_raw_labels(label_paths):\n",
        "    \"\"\"\n",
        "    Load raw labels from a list of text file paths.\n",
        "    Each file is expected to contain a sequence of numbers (as strings),\n",
        "    possibly with floating-point values. The function converts each token\n",
        "    to a float and then to an integer.\n",
        "    \"\"\"\n",
        "    raw_labels = []\n",
        "    for path in label_paths:\n",
        "        with open(path, \"r\") as f:\n",
        "            line = f.readline().strip()\n",
        "            # Convert each token to an integer via float conversion (handles tokens like \"0.8203125\")\n",
        "            seq = [int(float(x)) for x in line.split()]\n",
        "            raw_labels.append(seq)\n",
        "    return raw_labels\n",
        "\n",
        "# Load the raw label sequences for training and validation\n",
        "train_raw_labels = load_raw_labels(train_label_paths)\n",
        "valid_raw_labels = load_raw_labels(valid_label_paths)\n",
        "\n",
        "# Check the first few raw labels for inspection\n",
        "print(\"First 5 train raw labels:\", train_raw_labels[:5])\n",
        "print(\"First 5 valid raw labels:\", valid_raw_labels[:5])\n",
        "\n",
        "# Compute the length of each label sequence (i.e. number of tokens in each label)\n",
        "train_label_length = np.array([len(seq) for seq in train_raw_labels])\n",
        "valid_label_length = np.array([len(seq) for seq in valid_raw_labels])\n",
        "\n",
        "# Assume input_length is constant (e.g., the number of timesteps output by your model, here 256)\n",
        "train_input_length = np.full((len(train_raw_labels),), 256, dtype=np.int32)\n",
        "valid_input_length = np.full((len(valid_raw_labels),), 256, dtype=np.int32)\n",
        "\n",
        "# Print computed label lengths and input lengths for verification\n",
        "print(\"Train label lengths:\", train_label_length)\n",
        "print(\"Valid label lengths:\", valid_label_length)\n",
        "print(\"First 5 Train input lengths:\", train_input_length[:5])\n",
        "print(\"First 5 Valid input lengths:\", valid_input_length[:5])\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "D5MmL70tBNI7",
        "outputId": "f2f8f4bc-a9dd-4bc3-f6e4-357cbc5dfa49"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "First 5 train raw labels: [[50, 0, 0, 0, 0, 0, 0, 0, 0], [318, 0, 0, 0, 0, 0, 0, 0, 0], [318, 0, 0, 0, 0, 0, 0, 0, 0], [764, 0, 0, 0, 0, 0, 0, 0, 0], [324, 0, 0, 0, 0, 0, 0, 0, 0]]\n",
            "First 5 valid raw labels: [[628, 0, 0, 0, 0, 0, 0, 0, 0], [461, 0, 0, 0, 0, 0, 0, 0, 0], [667, 0, 0, 0, 0, 0, 0, 0, 0], [216, 0, 0, 0, 0, 0, 0, 0, 0], [216, 0, 0, 0, 0, 0, 0, 0, 0]]\n",
            "Train label lengths: [9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9]\n",
            "Valid label lengths: [9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9\n",
            " 9]\n",
            "First 5 Train input lengths: [256 256 256 256 256]\n",
            "First 5 Valid input lengths: [256 256 256 256 256]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "\n",
        "# Pad the training and validation raw labels with a padding value.\n",
        "# Use a value that is not a valid label (e.g., -1) so that the CTC loss ignores them.\n",
        "padded_train_labels = pad_sequences(train_raw_labels, padding='post', value=-1)\n",
        "padded_valid_labels = pad_sequences(valid_raw_labels, padding='post', value=-1)\n",
        "\n",
        "print(\"Padded train labels shape:\", padded_train_labels.shape)\n",
        "print(\"Padded valid labels shape:\", padded_valid_labels.shape)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5tE8Qj8eEmE0",
        "outputId": "6ef3a6ee-92f2-4855-a26e-5aa8bf78d892"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Padded train labels shape: (374, 9)\n",
            "Padded valid labels shape: (75, 9)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "\n",
        "# If train_images is a list, convert it to a NumPy array.\n",
        "if isinstance(train_images, list):\n",
        "    train_images = np.array(train_images)\n",
        "\n",
        "# Now you can safely check its shape.\n",
        "print(\"train_images shape:\", train_images.shape)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hZvWlBNFE95k",
        "outputId": "abefa8f5-607e-4b66-91d2-8432104806f0"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "train_images shape: (374,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"First train image element:\", train_images[0])\n",
        "print(\"Type of first element:\", type(train_images[0]))\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kJp7z2g6G1nr",
        "outputId": "5c2cf46c-79ca-432f-d645-096a73475744"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "First train image element: [[[0.42352942]\n",
            "  [0.43137255]\n",
            "  [0.42745098]\n",
            "  ...\n",
            "  [0.34901962]\n",
            "  [0.0627451 ]\n",
            "  [0.07843138]]\n",
            "\n",
            " [[0.3254902 ]\n",
            "  [0.32156864]\n",
            "  [0.34117648]\n",
            "  ...\n",
            "  [0.36862746]\n",
            "  [0.11372549]\n",
            "  [0.09803922]]\n",
            "\n",
            " [[0.3529412 ]\n",
            "  [0.34509805]\n",
            "  [0.3372549 ]\n",
            "  ...\n",
            "  [0.34117648]\n",
            "  [0.11372549]\n",
            "  [0.10980392]]\n",
            "\n",
            " ...\n",
            "\n",
            " [[0.3882353 ]\n",
            "  [0.3882353 ]\n",
            "  [0.39215687]\n",
            "  ...\n",
            "  [0.43137255]\n",
            "  [0.42352942]\n",
            "  [0.25882354]]\n",
            "\n",
            " [[0.41960785]\n",
            "  [0.42352942]\n",
            "  [0.43137255]\n",
            "  ...\n",
            "  [0.45882353]\n",
            "  [0.4509804 ]\n",
            "  [0.3254902 ]]\n",
            "\n",
            " [[0.23921569]\n",
            "  [0.24313726]\n",
            "  [0.24313726]\n",
            "  ...\n",
            "  [0.40392157]\n",
            "  [0.42352942]\n",
            "  [0.41568628]]]\n",
            "Type of first element: <class 'numpy.ndarray'>\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Convert every element to a string if needed.\n",
        "train_image_paths = [str(x) for x in train_images]\n"
      ],
      "metadata": {
        "id": "t3BDGi9XG8nK"
      },
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "\n",
        "# --- Part 1: Data Preprocessing ---\n",
        "\n",
        "# Assume `train_images` is already defined as a list of NumPy arrays.\n",
        "# Convert the list to a NumPy array.\n",
        "train_images = np.array(train_images)  # Now shape is something like (374, 32, 128, 1) or (374, 128, 32, 1)\n",
        "print(\"Initial train_images shape:\", train_images.shape)\n",
        "\n",
        "# If the images are in the shape (374, 32, 128, 1), transpose them to (374, 128, 32, 1)\n",
        "if train_images.shape[1:3] == (32, 128):\n",
        "    train_images = np.transpose(train_images, (0, 2, 1, 3))\n",
        "    print(\"Transposed train_images shape:\", train_images.shape)\n",
        "\n",
        "# Ensure images are in float32 format\n",
        "train_images = train_images.astype('float32')\n",
        "\n",
        "# Ensure your padded_train_labels is in int32 format.\n",
        "padded_train_labels = padded_train_labels.astype('int32')\n",
        "\n",
        "# Process train_label_length: make sure it's a NumPy array with shape (374, 1)\n",
        "train_label_length = np.array(train_label_length)\n",
        "train_label_length = np.squeeze(train_label_length)  # Remove extra dimensions if any\n",
        "if train_label_length.ndim == 1:\n",
        "    train_label_length = np.expand_dims(train_label_length, axis=-1)\n",
        "\n",
        "print(\"train_label_length shape:\", train_label_length.shape)  # Expected: (374, 1)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5thg-2VFJUoa",
        "outputId": "522fd2c6-f540-43f1-a1af-dd4ccb9457ab"
      },
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Initial train_images shape: (374, 128, 32, 1)\n",
            "train_label_length shape: (374, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# --- Part 2: Set train_input_length ---\n",
        "\n",
        "# Your CNN architecture produces an output sequence length of 64 (after pooling and reshaping).\n",
        "# Set train_input_length for every sample to 64.\n",
        "train_input_length = np.full((train_images.shape[0], 1), 64, dtype=np.int32)\n",
        "print(\"train_input_length shape:\", train_input_length.shape)  # Expected: (374, 1)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "t0jc6ojlJX_T",
        "outputId": "a22cf077-8a47-4ff3-ed5c-dc040a9cd895"
      },
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "train_input_length shape: (374, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow.keras import backend as K\n",
        "from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Dense, Lambda, Reshape\n",
        "from tensorflow.keras.models import Model\n",
        "\n",
        "# --- Part 3: Build the CTC Model ---\n",
        "\n",
        "def ctc_lambda_func(args):\n",
        "    y_pred, labels, input_length, label_length = args\n",
        "    return K.ctc_batch_cost(labels, y_pred, input_length, label_length)\n",
        "\n",
        "# Determine the correct number of classes based on your labels.\n",
        "max_label = np.max(padded_train_labels)\n",
        "# Set num_classes to max_label + 2. For example, if max_label is 318, then num_classes becomes 320.\n",
        "num_classes = int(max_label) + 2\n",
        "print(\"Max label value:\", max_label, \"-> Setting num_classes to:\", num_classes)\n",
        "\n",
        "# Define model inputs.\n",
        "input_img = Input(shape=(128, 32, 1), name='input_img')\n",
        "labels_input = Input(shape=(padded_train_labels.shape[1],), name='labels')\n",
        "input_length_input = Input(shape=(1,), name='input_length')\n",
        "label_length_input = Input(shape=(1,), name='label_length')\n",
        "\n",
        "# Build a simple CNN.\n",
        "x = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img)\n",
        "x = MaxPooling2D(pool_size=(2, 2))(x)\n",
        "# Reshape feature maps into a sequence; here the output becomes (None, 64, 512)\n",
        "x = Reshape(target_shape=(-1, x.shape[-1] * x.shape[-2]))(x)\n",
        "x = Dense(64, activation='relu')(x)\n",
        "\n",
        "# Use the updated num_classes here.\n",
        "y_pred = Dense(num_classes, activation='softmax', name='y_pred')(x)\n",
        "\n",
        "# Lambda layer for computing the CTC loss.\n",
        "loss_out = Lambda(ctc_lambda_func, output_shape=(1,), name='ctc')(\n",
        "    [y_pred, labels_input, input_length_input, label_length_input]\n",
        ")\n",
        "\n",
        "# Build and compile the model.\n",
        "model = Model(\n",
        "    inputs=[input_img, labels_input, input_length_input, label_length_input],\n",
        "    outputs=loss_out\n",
        ")\n",
        "model.compile(optimizer='adam', loss={'ctc': lambda y_true, y_pred: y_pred})\n",
        "model.summary()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 547
        },
        "id": "bd_QWMF8JaQL",
        "outputId": "f9fdc758-1839-43ae-a821-ef50ea42a5e2"
      },
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Max label value: 826 -> Setting num_classes to: 828\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1mModel: \"functional_8\"\u001b[0m\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"functional_8\"</span>\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┓\n",
              "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)             \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape          \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m       Param #\u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mConnected to          \u001b[0m\u001b[1m \u001b[0m┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━┩\n",
              "│ input_img (\u001b[38;5;33mInputLayer\u001b[0m)    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m1\u001b[0m)     │              \u001b[38;5;34m0\u001b[0m │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ conv2d_9 (\u001b[38;5;33mConv2D\u001b[0m)         │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m)    │            \u001b[38;5;34m320\u001b[0m │ input_img[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m]        │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ max_pooling2d_9           │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m32\u001b[0m)     │              \u001b[38;5;34m0\u001b[0m │ conv2d_9[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m]         │\n",
              "│ (\u001b[38;5;33mMaxPooling2D\u001b[0m)            │                        │                │                        │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ reshape_7 (\u001b[38;5;33mReshape\u001b[0m)       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m512\u001b[0m)        │              \u001b[38;5;34m0\u001b[0m │ max_pooling2d_9[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m]  │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ dense_6 (\u001b[38;5;33mDense\u001b[0m)           │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m64\u001b[0m)         │         \u001b[38;5;34m32,832\u001b[0m │ reshape_7[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m]        │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ y_pred (\u001b[38;5;33mDense\u001b[0m)            │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m828\u001b[0m)        │         \u001b[38;5;34m53,820\u001b[0m │ dense_6[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m]          │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ labels (\u001b[38;5;33mInputLayer\u001b[0m)       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m9\u001b[0m)              │              \u001b[38;5;34m0\u001b[0m │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ input_length (\u001b[38;5;33mInputLayer\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m)              │              \u001b[38;5;34m0\u001b[0m │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ label_length (\u001b[38;5;33mInputLayer\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m)              │              \u001b[38;5;34m0\u001b[0m │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ ctc (\u001b[38;5;33mLambda\u001b[0m)              │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m)              │              \u001b[38;5;34m0\u001b[0m │ y_pred[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m],          │\n",
              "│                           │                        │                │ labels[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m],          │\n",
              "│                           │                        │                │ input_length[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m],    │\n",
              "│                           │                        │                │ label_length[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m]     │\n",
              "└───────────────────────────┴────────────────────────┴────────────────┴────────────────────────┘\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┓\n",
              "┃<span style=\"font-weight: bold\"> Layer (type)              </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">        Param # </span>┃<span style=\"font-weight: bold\"> Connected to           </span>┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━┩\n",
              "│ input_img (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)     │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ conv2d_9 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv2D</span>)         │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)    │            <span style=\"color: #00af00; text-decoration-color: #00af00\">320</span> │ input_img[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>]        │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ max_pooling2d_9           │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">16</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)     │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ conv2d_9[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>]         │\n",
              "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">MaxPooling2D</span>)            │                        │                │                        │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ reshape_7 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Reshape</span>)       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">512</span>)        │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ max_pooling2d_9[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>]  │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ dense_6 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)           │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)         │         <span style=\"color: #00af00; text-decoration-color: #00af00\">32,832</span> │ reshape_7[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>]        │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ y_pred (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)            │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">828</span>)        │         <span style=\"color: #00af00; text-decoration-color: #00af00\">53,820</span> │ dense_6[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>]          │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ labels (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">9</span>)              │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ input_length (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>) │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)              │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ label_length (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>) │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)              │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ -                      │\n",
              "├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤\n",
              "│ ctc (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Lambda</span>)              │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)              │              <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │ y_pred[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>],          │\n",
              "│                           │                        │                │ labels[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>],          │\n",
              "│                           │                        │                │ input_length[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>],    │\n",
              "│                           │                        │                │ label_length[<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>][<span style=\"color: #00af00; text-decoration-color: #00af00\">0</span>]     │\n",
              "└───────────────────────────┴────────────────────────┴────────────────┴────────────────────────┘\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m86,972\u001b[0m (339.73 KB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">86,972</span> (339.73 KB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m86,972\u001b[0m (339.73 KB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">86,972</span> (339.73 KB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "\n",
        "# Assume train_label_length is a NumPy array with shape (374, 1)\n",
        "# and padded_train_labels has shape (374, 9)\n",
        "\n",
        "# Get indices for samples where label length > 0\n",
        "nonzero_indices = np.where(train_label_length[:, 0] > 0)[0]\n",
        "print(\"Number of samples before filtering:\", train_images.shape[0])\n",
        "print(\"Number of samples after filtering:\", nonzero_indices.shape[0])\n",
        "\n",
        "# Filter all corresponding arrays using these indices\n",
        "train_images = train_images[nonzero_indices]\n",
        "padded_train_labels = padded_train_labels[nonzero_indices]\n",
        "train_input_length = train_input_length[nonzero_indices]\n",
        "train_label_length = train_label_length[nonzero_indices]\n",
        "\n",
        "# Verify shapes\n",
        "print(\"Filtered train_images shape:\", train_images.shape)\n",
        "print(\"Filtered padded_train_labels shape:\", padded_train_labels.shape)\n",
        "print(\"Filtered train_input_length shape:\", train_input_length.shape)\n",
        "print(\"Filtered train_label_length shape:\", train_label_length.shape)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7C_m5JB7KJUj",
        "outputId": "54092e80-5691-4158-f622-e8a09b4a4ac1"
      },
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of samples before filtering: 374\n",
            "Number of samples after filtering: 373\n",
            "Filtered train_images shape: (373, 128, 32, 1)\n",
            "Filtered padded_train_labels shape: (373, 9)\n",
            "Filtered train_input_length shape: (373, 1)\n",
            "Filtered train_label_length shape: (373, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# --- Part 4: Train the Model ---\n",
        "\n",
        "# Create a dummy target; the Lambda layer outputs the loss directly.\n",
        "dummy_y = np.zeros((train_images.shape[0], 1), dtype='float32')\n",
        "\n",
        "history = model.fit(\n",
        "    x={\n",
        "        'input_img': train_images,\n",
        "        'labels': padded_train_labels,\n",
        "        'input_length': train_input_length,\n",
        "        'label_length': train_label_length\n",
        "    },\n",
        "    y=dummy_y,\n",
        "    batch_size=32,\n",
        "    epochs=100\n",
        ")\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Gt9gqu4sJdRQ",
        "outputId": "36fb9176-3ee2-460b-eb9f-b844e30905ac"
      },
      "execution_count": 54,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 147ms/step - loss: 5.8573\n",
            "Epoch 2/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 214ms/step - loss: 5.8754\n",
            "Epoch 3/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 165ms/step - loss: 5.9108\n",
            "Epoch 4/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 5.8511\n",
            "Epoch 5/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 129ms/step - loss: 5.7723\n",
            "Epoch 6/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 129ms/step - loss: 5.8041\n",
            "Epoch 7/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.8249\n",
            "Epoch 8/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 218ms/step - loss: 5.7516\n",
            "Epoch 9/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 134ms/step - loss: 5.7248\n",
            "Epoch 10/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 5.7591\n",
            "Epoch 11/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 5.6726\n",
            "Epoch 12/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 129ms/step - loss: 5.7124\n",
            "Epoch 13/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 221ms/step - loss: 5.6873\n",
            "Epoch 14/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 127ms/step - loss: 5.6455\n",
            "Epoch 15/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 127ms/step - loss: 5.6521\n",
            "Epoch 16/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 5.6237\n",
            "Epoch 17/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 149ms/step - loss: 5.6465\n",
            "Epoch 18/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 203ms/step - loss: 5.5825\n",
            "Epoch 19/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 5.6190\n",
            "Epoch 20/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 5.5594\n",
            "Epoch 21/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 131ms/step - loss: 5.5543\n",
            "Epoch 22/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 130ms/step - loss: 5.5193\n",
            "Epoch 23/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 199ms/step - loss: 5.5389\n",
            "Epoch 24/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 206ms/step - loss: 5.5433\n",
            "Epoch 25/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 5.4820\n",
            "Epoch 26/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.4932\n",
            "Epoch 27/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 130ms/step - loss: 5.4489\n",
            "Epoch 28/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.4618\n",
            "Epoch 29/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 129ms/step - loss: 5.4167\n",
            "Epoch 30/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 167ms/step - loss: 5.3407\n",
            "Epoch 31/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 199ms/step - loss: 5.3972\n",
            "Epoch 32/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.3528\n",
            "Epoch 33/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 5.4060\n",
            "Epoch 34/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 5.4397\n",
            "Epoch 35/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 131ms/step - loss: 5.4401\n",
            "Epoch 36/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 129ms/step - loss: 5.2967\n",
            "Epoch 37/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 216ms/step - loss: 5.3276\n",
            "Epoch 38/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 126ms/step - loss: 5.2025\n",
            "Epoch 39/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 127ms/step - loss: 5.2386\n",
            "Epoch 40/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 166ms/step - loss: 5.1725\n",
            "Epoch 41/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.2272\n",
            "Epoch 42/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 176ms/step - loss: 5.2439\n",
            "Epoch 43/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 217ms/step - loss: 5.1707\n",
            "Epoch 44/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 5.0857\n",
            "Epoch 45/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 5.1734\n",
            "Epoch 46/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 127ms/step - loss: 5.1736\n",
            "Epoch 47/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 5.0532\n",
            "Epoch 48/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 156ms/step - loss: 5.1607\n",
            "Epoch 49/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 207ms/step - loss: 5.0333\n",
            "Epoch 50/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 5.0156\n",
            "Epoch 51/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 131ms/step - loss: 5.0841\n",
            "Epoch 52/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.0254\n",
            "Epoch 53/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 5.0495\n",
            "Epoch 54/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 5.0210\n",
            "Epoch 55/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 166ms/step - loss: 4.9174\n",
            "Epoch 56/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 217ms/step - loss: 5.1006\n",
            "Epoch 57/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 126ms/step - loss: 4.9151\n",
            "Epoch 58/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 4.9797\n",
            "Epoch 59/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 4.8358\n",
            "Epoch 60/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 4.8935\n",
            "Epoch 61/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 4.8894\n",
            "Epoch 62/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 212ms/step - loss: 4.8366\n",
            "Epoch 63/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 4.8255\n",
            "Epoch 64/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 4.9026\n",
            "Epoch 65/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 4.8513\n",
            "Epoch 66/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 4.7973\n",
            "Epoch 67/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 125ms/step - loss: 4.6446\n",
            "Epoch 68/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 152ms/step - loss: 4.8009\n",
            "Epoch 69/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 220ms/step - loss: 4.6725\n",
            "Epoch 70/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 125ms/step - loss: 4.7113\n",
            "Epoch 71/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 128ms/step - loss: 4.6589\n",
            "Epoch 72/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 128ms/step - loss: 4.6732\n",
            "Epoch 73/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 135ms/step - loss: 4.6246\n",
            "Epoch 74/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 205ms/step - loss: 4.6895\n",
            "Epoch 75/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 131ms/step - loss: 4.7112\n",
            "Epoch 76/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 4.5776\n",
            "Epoch 77/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 131ms/step - loss: 4.7014\n",
            "Epoch 78/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 4.5358\n",
            "Epoch 79/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 131ms/step - loss: 4.5118\n",
            "Epoch 80/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 208ms/step - loss: 4.5437\n",
            "Epoch 81/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 4.5694\n",
            "Epoch 82/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 131ms/step - loss: 4.4462\n",
            "Epoch 83/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 127ms/step - loss: 4.6054\n",
            "Epoch 84/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 4.5105\n",
            "Epoch 85/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 4.4818\n",
            "Epoch 86/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 219ms/step - loss: 4.4525\n",
            "Epoch 87/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 178ms/step - loss: 4.3776\n",
            "Epoch 88/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 131ms/step - loss: 4.3235\n",
            "Epoch 89/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 131ms/step - loss: 4.3342\n",
            "Epoch 90/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 130ms/step - loss: 4.4222\n",
            "Epoch 91/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 4.3053\n",
            "Epoch 92/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 154ms/step - loss: 4.2582\n",
            "Epoch 93/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 203ms/step - loss: 4.2237\n",
            "Epoch 94/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 126ms/step - loss: 4.3418\n",
            "Epoch 95/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 131ms/step - loss: 4.2886\n",
            "Epoch 96/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 128ms/step - loss: 4.2118\n",
            "Epoch 97/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 130ms/step - loss: 4.2097\n",
            "Epoch 98/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 187ms/step - loss: 4.2595\n",
            "Epoch 99/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 191ms/step - loss: 4.1916\n",
            "Epoch 100/100\n",
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 129ms/step - loss: 4.1102\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras import backend as K\n",
        "\n",
        "# ================================\n",
        "# Part 1: Build the Inference Model\n",
        "# ================================\n",
        "# Assume your trained model was built using:\n",
        "#   input_img -> CNN + BiLSTM (or similar) -> Dense(softmax) = y_pred\n",
        "# For inference, we want to use only the image input and output y_pred.\n",
        "\n",
        "# If you defined your input layer as `input_img` and your final dense layer as `y_pred`\n",
        "# in your training model, you can build an inference model as follows:\n",
        "inference_model = tf.keras.models.Model(inputs=input_img, outputs=y_pred)\n",
        "print(\"Inference model summary:\")\n",
        "inference_model.summary()\n",
        "\n",
        "# ================================\n",
        "# Part 2: Decode the Predictions\n",
        "# ================================\n",
        "# Get the predictions for your evaluation dataset (e.g. train_images or test_images)\n",
        "predictions = inference_model.predict(train_images)  # shape: (num_samples, T, num_classes)\n",
        "T = predictions.shape[1]  # The sequence length (e.g. 64)\n",
        "\n",
        "# Use the built-in CTC decoder (greedy) to decode predictions.\n",
        "# Note: tf.keras.backend.ctc_decode returns a tuple:\n",
        "#       (decoded, log_probabilities)\n",
        "decoded_sequences, log_probabilities = K.ctc_decode(\n",
        "    predictions,\n",
        "    input_length=np.full((predictions.shape[0],), T, dtype=np.int32),\n",
        "    greedy=True\n",
        ")\n",
        "# decoded_sequences is a list of one element (dense tensor) with shape (num_samples, max_decoded_length)\n",
        "decoded_sequences = K.get_value(decoded_sequences[0])\n",
        "print(\"Decoded sequences shape:\", decoded_sequences.shape)\n",
        "\n",
        "# ================================\n",
        "# Part 3: Define a Helper to Remove Padding\n",
        "# ================================\n",
        "def remove_padding(seq, pad_value):\n",
        "    \"\"\"\n",
        "    Remove all occurrences of pad_value from the sequence.\n",
        "    For ground truth labels (padded with 0), use pad_value=0.\n",
        "    For decoded predictions (padded with -1 by default), use pad_value=-1.\n",
        "    \"\"\"\n",
        "    return [x for x in seq if x != pad_value]\n",
        "\n",
        "# ================================\n",
        "# Part 4: Compute Sequence Accuracy\n",
        "# ================================\n",
        "num_correct = 0\n",
        "num_samples = decoded_sequences.shape[0]\n",
        "\n",
        "for i in range(num_samples):\n",
        "    # Remove padding/blank tokens from the ground truth and predicted sequences.\n",
        "    # Adjust the pad_value if your ground truth padding is different.\n",
        "    true_seq = remove_padding(padded_train_labels[i], pad_value=0)\n",
        "    pred_seq = remove_padding(decoded_sequences[i], pad_value=-1)\n",
        "\n",
        "    # Compare the entire sequence.\n",
        "    if true_seq == pred_seq:\n",
        "        num_correct += 1\n",
        "\n",
        "sequence_accuracy = num_correct / num_samples\n",
        "print(\"Sequence Accuracy: {:.2%}\".format(sequence_accuracy))\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 407
        },
        "id": "OO3B3nM9LSVG",
        "outputId": "d8f146fc-41a0-453d-f357-023fa3a4e980"
      },
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Inference model summary:\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1mModel: \"functional_10\"\u001b[0m\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"functional_10\"</span>\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                        \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape               \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m        Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ input_img (\u001b[38;5;33mInputLayer\u001b[0m)               │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m1\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_9 (\u001b[38;5;33mConv2D\u001b[0m)                    │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m128\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m)         │             \u001b[38;5;34m320\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_9 (\u001b[38;5;33mMaxPooling2D\u001b[0m)       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m32\u001b[0m)          │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ reshape_7 (\u001b[38;5;33mReshape\u001b[0m)                  │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m512\u001b[0m)             │               \u001b[38;5;34m0\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense_6 (\u001b[38;5;33mDense\u001b[0m)                      │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m64\u001b[0m)              │          \u001b[38;5;34m32,832\u001b[0m │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ y_pred (\u001b[38;5;33mDense\u001b[0m)                       │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m, \u001b[38;5;34m828\u001b[0m)             │          \u001b[38;5;34m53,820\u001b[0m │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃<span style=\"font-weight: bold\"> Layer (type)                         </span>┃<span style=\"font-weight: bold\"> Output Shape                </span>┃<span style=\"font-weight: bold\">         Param # </span>┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ input_img (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">InputLayer</span>)               │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ conv2d_9 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv2D</span>)                    │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">128</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)         │             <span style=\"color: #00af00; text-decoration-color: #00af00\">320</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ max_pooling2d_9 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">MaxPooling2D</span>)       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">16</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">32</span>)          │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ reshape_7 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Reshape</span>)                  │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">512</span>)             │               <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense_6 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                      │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>)              │          <span style=\"color: #00af00; text-decoration-color: #00af00\">32,832</span> │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ y_pred (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                       │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">64</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">828</span>)             │          <span style=\"color: #00af00; text-decoration-color: #00af00\">53,820</span> │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m86,972\u001b[0m (339.73 KB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">86,972</span> (339.73 KB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m86,972\u001b[0m (339.73 KB)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">86,972</span> (339.73 KB)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
            ],
            "text/html": [
              "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
              "</pre>\n"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1m12/12\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 38ms/step\n",
            "Decoded sequences shape: (373, 64)\n",
            "Sequence Accuracy: 0.00%\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Plot training loss over epochs\n",
        "plt.figure(figsize=(10, 5))\n",
        "plt.plot(history.history['loss'], marker='o', label='Training Loss')\n",
        "plt.title('CTC Loss over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "plt.grid(True)\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 371
        },
        "id": "OFKC2EDlMisz",
        "outputId": "b205d2bf-1a55-4d98-8e3a-044d090450e7"
      },
      "execution_count": 56,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1000x500 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Example: Assume you have a list `cer_values` calculated for each epoch.\n",
        "cer_values = [0.75, 0.65, 0.60, 0.55, 0.50]  # Replace with your actual values\n",
        "\n",
        "plt.figure(figsize=(10, 5))\n",
        "plt.plot(cer_values, marker='o', color='red', label='Character Error Rate')\n",
        "plt.title('CER Over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('CER')\n",
        "plt.legend()\n",
        "plt.grid(True)\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 370
        },
        "id": "2szNz5YJMu4A",
        "outputId": "e4fe8e1e-1088-4503-8aaa-a7447cdac779"
      },
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1000x500 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1cAAAHWCAYAAACbsXOkAAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjEwLjAsIGh0dHBzOi8vbWF0cGxvdGxpYi5vcmcvlHJYcgAAAAlwSFlzAAAPYQAAD2EBqD+naQAAdJ1JREFUeJzt3Xd4FGXbxuHfpidAaIEkhBK6dCRIpEkngCLFAopSVJqgFKUpgqCCFGmKikgTCwgqitQYmvQSQUA6hN4FAgGSkMz3x3zsa0iABLKZlOs8jj1ednZ29t6bMS9XnpnnsRmGYSAiIiIiIiIPxcnqAkRERERERDIDhSsREREREZFUoHAlIiIiIiKSChSuREREREREUoHClYiIiIiISCpQuBIREREREUkFClciIiIiIiKpQOFKREREREQkFShciYiIiIiIpAKFKxERkSxm5syZ2Gw2tm7danUpIiKZisKViIjYHTp0iK5du1KsWDE8PDzw9vamZs2aTJw4kRs3btj3CwwMxGazJflo0qSJfb/3338/wWuurq4EBgby5ptvcvny5RTV9vvvv9OkSRPy5s2Lh4cHpUqV4u233+bixYup9fVTze3wcrfHxo0brS5RREQcwMXqAkREJH1YtGgRzz33HO7u7rRv357y5csTExPD2rVr6devH7t37+arr76y71+5cmXeeuutRMcpUKBAom1ffPEF2bNnJyoqirCwMD799FPCw8NZu3Ztsmp7++23+eSTT6hUqRIDBgwgT548hIeH89lnnzFnzhzCwsIoXbr0g395Bxk+fDhFixZNtL1EiRIWVCMiIo6mcCUiIhw5coS2bdtSpEgRVqxYgb+/v/21Hj16cPDgQRYtWpTgPQEBAbz00kvJOv6zzz6Lj48PAF27dqVt27bMnTuXzZs3U61atXu+94cffuCTTz6hTZs2fPfddzg7O9tf69ixI/Xq1eO5554jPDwcF5e0+7+1qKgosmXLds99mjZtStWqVdOoIhERsZouCxQREUaPHs21a9eYNm1agmB1W4kSJejVq1eqfV7t2rUB8zLE+xk2bBi5c+fmq6++ShCsAKpVq8aAAQPYuXMn8+fPB6Bnz55kz56d69evJzrWCy+8gJ+fH3FxcfZtS5YsoXbt2mTLlo0cOXLw5JNPsnv37gTv69ixI9mzZ+fQoUM0a9aMHDly0K5duxR/7ztFRERgs9kYO3Ys48ePp0iRInh6elKnTh127dqVaP8VK1bYa82VKxctWrRgz549ifY7efIkr776KgUKFMDd3Z2iRYvSvXt3YmJiEuwXHR1N3759yZcvH9myZaNVq1acP38+wT5bt24lJCQEHx8fPD09KVq0KK+88spDf3cRkcxII1ciIsLChQspVqwYNWrUSPZ7YmNjuXDhQqLt2bJlw9PT857vjYiIACB37tz33O/AgQPs27ePjh074u3tneQ+7du3Z+jQofz++++0bduWNm3aMHnyZPtljrddv36dhQsX0rFjR3tImz17Nh06dCAkJIRRo0Zx/fp1vvjiC2rVqsVff/1FYGCg/f23bt0iJCSEWrVqMXbsWLy8vO5ZO8CVK1cS9chms5E3b94E27755huuXr1Kjx49uHnzJhMnTqR+/frs3LkTX19fAP744w+aNm1KsWLFeP/997lx4waffvopNWvWJDw83F7rqVOnqFatGpcvX6ZLly488sgjnDx5kvnz53P9+nXc3Nzsn/vGG2+QO3duhg4dSkREBBMmTKBnz57MnTsXgHPnztG4cWPy5cvHwIEDyZUrFxEREfz888/3/e4iIlmSISIiWdqVK1cMwGjRokWy31OkSBEDSPIxcuRI+35Dhw41AGPfvn3G+fPnjYiICGP69OmGp6enkS9fPiMqKuqen7NgwQIDMMaPH3/P/by9vY0qVaoYhmEY8fHxRkBAgPHMM88k2OfHH380AGPNmjWGYRjG1atXjVy5chmdO3dOsN+ZM2eMnDlzJtjeoUMHAzAGDhx4394YhmHMmDHjrv1xd3e373fkyBEDMDw9PY0TJ07Yt2/atMkAjD59+ti3Va5c2cifP79x8eJF+7YdO3YYTk5ORvv27e3b2rdvbzg5ORlbtmxJVFd8fHyC+ho2bGjfZhiG0adPH8PZ2dm4fPmyYRiG8csvvxhAkscSEZHENHIlIpLFRUZGApAjR44UvS84OJgPP/ww0faSJUsm2nbnZBMVKlRgxowZ9x39uXr1arJqy5Ejh/172Gw2nnvuOaZMmcK1a9fInj07AHPnziUgIIBatWoBEBoayuXLl3nhhRcSjC45OzsTHBzMypUrE31O9+7d71nHnSZPnkypUqUSbLvz0kaAli1bEhAQYH9erVo1goODWbx4MePGjeP06dNs376d/v37kydPHvt+FStWpFGjRixevBiA+Ph4FixYQPPmzZO818tmsyV43qVLlwTbateuzfjx4zl69CgVK1YkV65cgDlTY6VKlXB1dU3R9xcRyWoUrkREsrjbl9vdDjLJ5ePjQ8OGDZO1708//YS3tzfnz59n0qRJHDly5L6XDsL/QtX9art69Sr58+e3P2/Tpg0TJkzgt99+48UXX+TatWssXryYrl272sPEgQMHAKhfv36Sx7zzMkQXFxcKFix435r/q1q1asma0CKpQFqqVCl+/PFHAI4ePQokDqkAZcqUYdmyZURFRXHt2jUiIyMpX758suorXLhwgue3L9O8dOkSAHXq1OGZZ55h2LBhjB8/nrp169KyZUtefPFF3N3dk/UZIiJZicKViEgW5+3tTYECBZKcQCG1PPHEE/bZAps3b06FChVo164d27Ztw8np7nMrlSlTBoC///77rvscPXqUyMhIypYta9/2+OOPExgYyI8//siLL77IwoULuXHjBm3atLHvEx8fD5j3Xfn5+SU67p0zD7q7u9+z1owoqVE0AMMwAHOka/78+WzcuJGFCxeybNkyXnnlFT755BM2btxoHxUUERFT5vp/CREReSBPPfUUhw4dYsOGDQ7/rOzZszN06FC2b99uH5m5m1KlSlGqVCkWLFhw19Grb775BjC/w389//zzLF26lMjISObOnUtgYCCPP/64/fXixYsDkD9/fho2bJjoUbdu3Yf4lilzexTtv/bv32+fpKJIkSIA7Nu3L9F+e/fuxcfHh2zZspEvXz68vb1TPSg//vjjfPTRR2zdupXvvvuO3bt3M2fOnFT9DBGRzEDhSkRE6N+/P9myZeO1117j7NmziV4/dOgQEydOTLXPa9euHQULFmTUqFH33XfIkCFcunSJbt26JZhCHWDbtm2MGjWK8uXL88wzzyR4rU2bNkRHRzNr1iyWLl3K888/n+D1kJAQvL29GTFiBLGxsYk+984pyR1pwYIFnDx50v588+bNbNq0iaZNmwLg7+9P5cqVmTVrFpcvX7bvt2vXLpYvX06zZs0AcHJyomXLlixcuJCtW7cm+pzbI1LJdenSpUTvqVy5MmBO4y4iIgnpskAREaF48eJ8//33tGnThjJlytC+fXvKly9PTEwM69evZ968eXTs2DHBe06ePMm3336b6FjZs2enZcuW9/w8V1dXevXqRb9+/Vi6dClNmjS5677t2rVjy5YtTJw4kX/++Yd27dqRO3duwsPDmT59Onnz5mX+/PmJJluoUqUKJUqU4N133yU6OjrBJYFgXg75xRdf8PLLL1OlShXatm1Lvnz5OHbsGIsWLaJmzZp89tln927cfSxZsoS9e/cm2l6jRg2KFStmf16iRAlq1apF9+7diY6OZsKECeTNm5f+/fvb9xkzZgxNmzalevXqvPrqq/ap2HPmzMn7779v32/EiBEsX76cOnXq0KVLF8qUKcPp06eZN28ea9eutU9SkRyzZs3i888/p1WrVhQvXpyrV68ydepUvL297YFORET+w+LZCkVEJB3Zv3+/0blzZyMwMNBwc3MzcuTIYdSsWdP49NNPjZs3b9r3u9dU7EWKFLHvd3sq9vPnzyf6rCtXrhg5c+Y06tSpk6zaFixYYDRq1MjInTu34e7ubpQoUcJ46623kjz2be+++64BGCVKlLjrPitXrjRCQkKMnDlzGh4eHkbx4sWNjh07Glu3brXv06FDByNbtmzJqtMw7j0VO2DMmDHDMIz/TcU+ZswY45NPPjEKFSpkuLu7G7Vr1zZ27NiR6Lh//PGHUbNmTcPT09Pw9vY2mjdvbvzzzz+J9jt69KjRvn17I1++fIa7u7tRrFgxo0ePHkZ0dHSC+u6cYn3lypUGYKxcudIwDMMIDw83XnjhBaNw4cKGu7u7kT9/fuOpp55K0BsREfkfm2Gk8BoBERERSRUREREULVqUMWPG8Pbbb1tdjoiIPCTdcyUiIiIiIpIKFK5ERERERERSgcKViIiIiIhIKtA9VyIiIiIiIqlAI1ciIiIiIiKpQOFKREREREQkFWgR4STEx8dz6tQpcuTIgc1ms7ocERERERGxiGEYXL16lQIFCuDkdO+xKYWrJJw6dYpChQpZXYaIiIiIiKQTx48fp2DBgvfcR+EqCTly5ADMBnp7e1taS2xsLMuXL6dx48a4urpaWktmpP46lvrrWOqv46nHjqX+Opb661jqr2Olp/5GRkZSqFAhe0a4F4WrJNy+FNDb2ztdhCsvLy+8vb0tP7EyI/XXsdRfx1J/HU89diz117HUX8dSfx0rPfY3ObcLaUILERERERGRVKBwJSIiIiIikgoUrkRERERERFKB7rkSEREREQzD4NatW8TFxVldSoYQGxuLi4sLN2/eVM8cIC376+zsjIuLS6oswaRwJSIiIpLFxcTEcPr0aa5fv251KRmGYRj4+flx/PhxrYvqAGndXy8vL/z9/XFzc3uo4yhciYiIiGRh8fHxHDlyBGdnZwoUKICbm5vCQjLEx8dz7do1smfPft+FZSXl0qq/hmEQExPD+fPnOXLkCCVLlnyoz1O4EhEREcnCYmJiiI+Pp1ChQnh5eVldToYRHx9PTEwMHh4eClcOkJb99fT0xNXVlaNHj9o/80HpTBARERERBQTJ0lLr/Nd/RSIiIiIiIqlA4So9i4vDtno1AWvWYFu9GjQTjYiIiIhIupUuwtXkyZMJDAzEw8OD4OBgNm/efNd969ati81mS/R48skn7ft07Ngx0etNmjRJi6+Sen7+GQIDcWnUiKrjxuHSqBEEBprbRURERNKjuDhYtQp++MH833Tyi2GbzcaCBQusLkOyAMvD1dy5c+nbty9Dhw4lPDycSpUqERISwrlz55Lc/+eff+b06dP2x65du3B2dua5555LsF+TJk0S7PfDDz+kxddJHT//DM8+CydOJNx+8qS5XQFLRERE0pv//8Uw9erBiy+a/5sGvxg+c+YMb7zxBsWKFcPd3Z1ChQrRvHlzwsLCHPq5qaVjx460bNnSYcefOXNmkgMTDzNpQ2r5bz3e3t489thj/Prrryk6RkREBDabje3btzumyBSyPFyNGzeOzp0706lTJ8qWLcuXX36Jl5cX06dPT3L/PHny4OfnZ3+Ehobi5eWVKFy5u7sn2C937txp8XUeXlwc9OoFhpH4tdvbevdON78JEhEREbHqF8MREREEBQWxYsUKxowZw86dO1m6dCn16tWjR48eDvnM22JiYhx6/JS6Vz3e3t4JBh1Onz7N0aNHU3Ss24tMp9T93jdjxgxOnz7N1q1bqVmzJs8++yw7d+5M8eekF5ZOxR4TE8O2bdsYNGiQfZuTkxMNGzZkw4YNyTrGtGnTaNu2LdmyZUuwfdWqVeTPn5/cuXNTv359PvzwQ/LmzZvkMaKjo4mOjrY/j4yMBMyVoWNjY1P6tR6KbfVqXO78wfRfhgHHj3Nr5UqMOnXSrrBM6vbfb1r/PWcV6q9jqb+Opx47lvrrWMntb2xsLIZhEB8fT3x8vLnRMCC5CwrHxWF74w0wDBKtjmUYGDYbvPkmRv364Ox8/+N5eUEy19nq3r07NpuNjRs3Jvi3YJkyZejYseP/vg9w7tw5WrZsyfLlywkICGDMmDE8/fTT//8V4ujatSsrV67kzJkzFC5cmO7du/Pmm2/a39+pUycuX77MY489xueff467uzt//fUXs2fP5tNPP2Xfvn1ky5aNevXqMX78ePLnz29/7+7duxk4cCB//vknhmFQuXJlpk+fzrfffsusWbMA7GuLhYWFUbduXY4fP87bb79NaGgoTk5O1KpViwkTJhAYGHjXeg4dOpSoR/Hx8dhstgT1/Pc1gPr161OuXDlcXFz47rvvqFChAu+99x4NGjTg999/Z8iQIfbgWr16dfr378/cuXOJjIykatWqfPLJJzz22GOA+W/wpN5Xt27dJP8Ovb29yZ8/P/nz52fYsGFMnDiRFStWULZsWQCWLFnCyJEj7VerPf7440yYMIHixYsDULRoUQAeffRRAOrUqcOKFSsA+Prrrxk/fjxHjhwhMDCQN954g+7duydZR3x8PIZhEBsbi/Md52lKfkZZGq4uXLhAXFwcvr6+Cbb7+vqyd+/e+75/8+bN7Nq1i2nTpiXY3qRJE1q3bk3RokU5dOgQ77zzDk2bNmXDhg2JmgUwcuRIhg0blmj78uXL03y9h4A1a6iajP22L1nCyagoh9eTVYSGhlpdQqam/jqW+ut46rFjqb+Odb/+uri44Ofnx7Vr1/43YhEVRa6CBVPl822GASdPYkvmVUSXT5yAO35pnpRLly6xbNkyBg8eTFxcnP2X47c5OTkl2DZs2DCGDRvGkCFD+Oqrr3j55Zf5+++/yZ07N7GxseTLl4/p06eTJ08eNm3aRJ8+fciZMyetWrUCzH9gr1ixAk9PT3766Sf7ca9evcqAAQMoWbIk58+f59133+Xll19m3rx5AJw6dYo6depQq1Ytfv31V3LkyMGmTZu4fPkynTt3ZufOnURGRjJ58mQAcufOzcWLFwkJCeGxxx5j0aJFuLi4MHbsWJo0acLatWtxc3NLsp47ewBw8+ZNDMNI8rXbbt26xTfffEOnTp1YsmQJAGfPngVgwIABfPDBBwQGBpIrVy769OnDb7/9xuTJkylUqBCTJk2iSZMmhIeHkzt3bq7/fyi/8313+/wbN24QGRnJrVu3mDp1KmCG3atXrwJw8eJFunbtSrly5YiKimLEiBG0bNmSP//8EycnJ8LCwmjQoAELFizgkUcewc3NjcjISH788UeGDh3K6NGjqVixIn///Te9evXCycmJF154IVEdMTEx3LhxgzVr1iQaabue3F80kMEXEZ42bRoVKlSgWrVqCba3bdvW/ucKFSpQsWJFihcvbk/Sdxo0aBB9+/a1P4+MjKRQoUI0btwYb29vx32BJNiyZYNx4+67X+WmTamkkauHFhsbS2hoKI0aNcLV1dXqcjId9dex1F/HU48dS/11rOT29+bNmxw/fpzs2bP/7z6c5IwwOYi3t3eywtXevXsxDINKlSol699rnTp14pVXXgFgzJgxTJkyhT179tgnPRs5cqR93woVKrBjxw5+//13OnToAICrqyvZsmVj5syZuLm5YRgGV69etY+e3ZYzZ06Cg4NxcnIie/bsjBo1ily5cjFv3jz730OVKlXs++fIkYO4uDhKlixp3/btt98C/7tfCmD27NnkyZOH8PBwGjdunKieu/Hw8CAyMpKCd4TlWrVqsXjxYsAM2CVLlmTChAn211etWgXABx98QIsWLQCIiopi+vTpTJ8+nWeeeQYwL+srVqwY8+bN4+2337YPTPz3fffy2muv4ezszI0bN4iPjycwMJD27duTI0cOrl69Srt27RL0d9asWfj6+nLixAnKly9vH8krVKhQgh6OHj2asWPH2oNUhQoViIiIYPbs2XTt2jVRHTdv3sTT05Mnnngi0f1o9wqmd7I0XPn4+ODs7GxPxredPXsWPz+/e743KiqKOXPmMHz48Pt+TrFixfDx8eHgwYNJhit3d3fc3d0TbXd1dU37H/b16kHBguY1yknddwXg7o5LxYqg/yNKNZb8XWch6q9jqb+Opx47lvrrWPfrb1xcHDabDScnp/8tpJo9O1y7lrwPWLMGmjW7/36LF8MTT9x3N6dkXhZ4+x/cCeq+h0qVKtn3y5EjB97e3ly4cMG+bfLkyUyfPp1jx45x48YNYmJiqFy5sv11m81GhQoV7P/wvn1JXXh4OMOHD2fHjh1cunTJvv3EiROULVuWHTt2ULt27ST/rXn7uLf7f9vOnTs5ePAgOXPmTLDvzZs3OXLkCE5OTonquRsnJydy5MhBeHh4gu2enp4JPjMoKCjB89t/rlatmv3PR44cITY2ltq1a9u3ubu7U61aNfbu3Zvg7+K/77uX8ePH07BhQw4fPkyfPn2YNGkSPj4+9j4ePHiQ999/n02bNnHhwoUE/a1YsaL9M/772VFRURw6dIjOnTsnCFK3bt0iZ86cSdZ1u6dJ/feSkp9PloYrNzc3goKCCAsLs8+SEh8fT1hYGD179rzne+fNm0d0dDQvvfTSfT/nxIkTXLx4EX9//9Qo27GcnWHiRPPmT5st6YAVHQ3BwfDLL1C5cpqXKCIiIpmczZas0SMAGje+9y+GbTbz9caNU3VErGTJkthstmTdSgKJ/4Fss9ns/1CfM2cOb7/9Np988gnVq1cnR44cjBkzhk2bNiV4z533+EdFRdG0aVNCQkL47rvvyJcvH8eOHSMkJMR+iaWnp2eKv9u1a9cICgriu+++S/Ravnz57lrP3Tg5OVGiRIl77nO3YyX3Mx70fX5+fpQoUYISJUowY8YMmjVrxj///IOPjw8ALVq0oEiRIkydOpUCBQoQHx9P+fLl7zmBx7X//8XA1KlTCQ4OTvBaUrcIpSbLZwvs27cvU6dOZdasWezZs4fu3bsTFRVFp06dAGjfvn2CCS9umzZtGi1btkw0ScW1a9fo168fGzduJCIigrCwMFq0aEGJEiUICQlJk+/00Fq3hvnzISAg4fZCheCTT6BYMYiIgBo14PvvLSlRREREBPjfL4Yh8YjT7ecTJqT6pYZ58uQhJCSEyZMnE5XEfeiXL19O9rHWrVtHjRo1eP3113n00UcpUaJEkpND3OnAgQNcvHiRjz/+mNq1a/PII48kWk6oYsWK/Pnnn3edFMHNzY24O2aBrlKlCgcOHCB//vz24HH7cedoVloqXrw4bm5urFu3zr4tNjaWLVu22CegeBjVqlUjKCiIjz76CIB///2Xffv2MXjwYBo0aECZMmW4dOlSgvfcviTyvz309fWlQIECHD58OFH/bk+A4SiWh6s2bdowduxYhgwZQuXKldm+fTtLly61T3Jx7NgxTp8+neA9+/btY+3atbz66quJjufs7Mzff//N008/TalSpXj11VcJCgrizz//vOtwbLrUujVERHArNJStfftyKzQUjhyBvn1hyxZo0gRu3IB27cxtDzA1poiIiEiquNsvhgsWNLe3bu2Qj508eTJxcXFUq1aNn376iQMHDrBnzx4mTZpE9erVk32ckiVLsnXrVpYtW8b+/ft577332LJly33fV7BgQdzc3Pj00085fPgwv/32Gx988EGCfXr27ElkZCRt27Zl69atHDhwgNmzZ7Nv3z4AAgMD+fvvv9m3bx8XLlwgNjaWdu3a4ePjQ4sWLfjzzz85cuQIq1at4s033+TEvWaVvgvDMDhz5kyix39nU0yObNmy0b17d/r168fSpUv5559/6Ny5M9evX0/y3+UPonfv3kyZMoWTJ0+SK1cu8ubNy1dffcXBgwdZsWJFgnkSAPLnz4+npydLly7l7NmzXLlyBTAnMBk5ciSTJk1i//797Ny5kxkzZjAuGXMbPAzLwxWYJ93Ro0eJjo5m06ZNCYbvVq1axcyZMxPsX7p0aQzDoFGjRomO5enpybJlyzh37hwxMTFERETw1VdfJZqRMENwdsaoU4eTTzxhTrt++zc+efLA77/DO++Yz8ePh0aN4C4LL4uIiIg43P//YpiVK80ra1auNH8x7KBgBeZ99eHh4dSrV4+33nqL8uXL06hRI8LCwvjiiy+SfZyuXbvSunVr2rRpQ3BwMBcvXuT111+/7/t8fHyYPn068+bNo2zZsnz88ceMHTs2wT558+ZlxYoVXLt2jTp16hAUFMTUqVPtlyl27tyZ0qVLU7VqVfLly8e6devw8vJizZo1FC5cmNatW1OmTBleffVVbt68+UCTrUVGRuLv75/ocecoW3J8/PHHPPPMM7z88stUqVKFgwcPsmzZslRbU7ZJkyYULVqUESNG4OTkxPfff8+2bdsoX748ffr0YcyYMQn2d3FxYdKkSUyZMoUCBQrYJ9F47bXX+Prrr5kxYwYVKlSgTp06zJw50+EjVzbDuNusCVlXZGQkOXPm5MqVK2k+W+CdYmNjWbx4Mc2aNUv6Zrqff4YOHcybTgsVMp9XTc5k7gLJ6K88FPXXsdRfx1OPHUv9dazk9vf2JAlFixa97+QI8j/x8fFERkbi7e2drIkbJGXSur/3+u8gJdlAZ0JG17o1bNoEpUrB8eNQqxbcMdInIiIiIiKOp3CVGZQtC5s3Q/Pm5kyCnTpBjx5wj1lUREREREQkdSlcZRY5c8KCBTBsmPn888+hfn04c8bSskREREREsgqFq8zEyQmGDIGFC82wtW4dVKkCGzZYXZmIiIiISKancJUZPfWUOV172bJw+jTUqQNTpiS9sJ+IiIgI5nTdIllVap3/CleZVcmS5kQXzz4LsbHQrRt07gw3b1pdmYiIiKQjt2cSvH79usWViFjn9vn/sDOXuqRGMZJOZc8OP/4Io0eba2JNmwY7d8JPP5mL+omIiEiW5+zsTK5cuexrHnl5eWGz2SyuKv2Lj48nJiaGmzdvaip2B0ir/hqGwfXr1zl37hy5cuXC+fa6sg9I4Sqzs9lgwAB49FFo29acVTAoCObNgyeesLo6ERERSQf8/PwAHmhR2azKMAxu3LiBp6enwqgDpHV/c+XKZf/v4GEoXGUVjRvD1q3mulg7dkCDBvDJJ/DGG2YAExERkSzLZrPh7+9P/vz5iY2NtbqcDCE2NpY1a9bwxBNPaBFsB0jL/rq6uj70iNVtCldZSbFisH69ee/V999Dr15m4JoyBTw9ra5ORERELObs7Jxq/8jM7Jydnbl16xYeHh4KVw6QUfurC0SzGi8v+PZbGDcOnJ1h9myoWRMiIqyuTEREREQkQ1O4yopsNujTB0JDwccH/voLqlaFsDCrKxMRERERybAUrrKyevVg2zZzgouLF837ssaO1XpYIiIiIiIPQOEqqytcGP78Ezp2hPh46NcPXngBoqKsrkxEREREJENRuBJzMovp02HyZHBxgblzoXp1OHTI6spERERERDIMhSsx2Wzw+uuwciX4+pqLDVetCkuXWl2ZiIiIiEiGoHAlCdWqZd6H9fjjcPkyNGsGI0boPiwRERERkftQuJLEAgJg1Sro2tUMVe++C888A1evWl2ZiIiIiEi6pXAlSXN3hy+/hKlTwc0NfvkFqlWDffusrkxEREREJF1SuJJ7e+01WLPGHM3auxceewx+/dXqqkRERERE0h2FK7m/4GDzPqzatc1LA1u2hCFDzKnbRUREREQEULiS5PL1hbAwePNN8/kHH8DTT5uTXoiIiIiIiMKVpICrK0ycCN98Ax4esGiReZng7t1WVyYiIiIiYjmFK0m5l1+GdeugcGE4eNC8bHD+fKurEhERERGxlMKVPJgqVcz7sBo0gKgoeO45GDgQ4uKsrkxERERExBIKV/LgfHxg6VJ4+23z+ahR5qLDFy9aW5eIiIiIiAUUruThuLjAmDHwww/g5QXLl0PVqrB9u9WViYiIiIikKYUrSR1t28KGDVCsGEREQI0a8P33VlclIiIiIpJmFK4k9VSsCFu2QJMmcOMGtGsHffvCrVtWVyYiIiIi4nAKV5K68uSB33+Hd94xn48fD40awblz1tYlIiIiIuJgCleS+pyd4aOP4OefIXt2WLXKvA9r61arKxMRERERcRiFK3GcVq1g82YoXRqOH4datWDGDKurEhERERFxCIUrcawyZWDTJnj6aYiOhldegR49ICbG6spERERERFKVwpU4Xs6c8MsvMHw42Gzw+edQvz6cPm11ZSIiIiIiqUbhStKGkxO89x4sXGiGrXXrICjInL5dRERERCQTULiStPXkk+Z07WXLmiNXderAlClgGFZXJiIiIiLyUBSuJO2VLGneh/XssxAbC926QefOcPOm1ZWJiIiIiDwwhSuxRvbs8OOP8PHH5iWD06aZo1gnTlhdmYiIiIjIA1G4EuvYbDBgACxZArlzm9O2BwXBmjVWVyYiIiIikmIKV2K9xo3NBYYrVYJz56BBA5g0SfdhiYiIiEiGonAl6UOxYrB+Pbz4Ity6Bb16QYcOcOOG1ZWJiIiIiCSLwpWkH15e8O23MG4cODvD7NlQsyZERFhdmYiIiIjIfSlcSfpis0GfPhAaCj4+8NdfULUqhIVZXZmIiIiIyD0pXEn6VK8ebNtmTnBx8aJ5X9bYsboPS0RERETSLYUrSb8KF4Y//4SOHSE+Hvr1gxdegKgoqysTEREREUlE4UrSN09PmD4dPv8cXFxg7lyoXh0OHbK6MhERERGRBBSuJP2z2aB7d1i5Evz8YOdO8z6sJUusrkxERERExE7hSjKOWrXM+7CqV4fLl+HJJ+Gjj8xLBkVERERELKZwJRlLgQLmCFbXrubkFoMHwzPPQGSk1ZWJiIiISBancCUZj7s7fPklTJ0Kbm6wYAEEB8O+fVZXJiIiIiJZmMKVZFyvvQZr1kBAAOzdC489Br/+anVVIiIiIpJFKVxJxhYcbN6HVbs2XL0KLVvCkCG6D0tERERE0pzClWR8vr4QFgZvvmk+/+ADePppc9ILEREREZE0onAlmYOrK0ycCN98Ax4esGiReZng7t1WVyYiIiIiWYTClWQuL78M69ZB4cJw8KB52eD8+VZXJSIiIiJZgMKVZD5Vqpj3YTVoAFFR8NxzMHAgxMVZXZmIiIiIZGIKV5I5+fjA0qXw9tvm81GjoFkzuHjR2rpEREREJNNSuJLMy8UFxoyBH34ALy9YvhyqVoXt262uTEREREQyoXQRriZPnkxgYCAeHh4EBwezefPmu+5bt25dbDZboseTTz5p38cwDIYMGYK/vz+enp40bNiQAwcOpMVXkfSobVvYsAGKFYOICKhRA777zuqqRERERCSTsTxczZ07l759+zJ06FDCw8OpVKkSISEhnDt3Lsn9f/75Z06fPm1/7Nq1C2dnZ5577jn7PqNHj2bSpEl8+eWXbNq0iWzZshESEsLNmzfT6mtJelOxImzdCk2bwo0b8NJL0KcPxMZaXZmIiIiIZBKWh6tx48bRuXNnOnXqRNmyZfnyyy/x8vJi+vTpSe6fJ08e/Pz87I/Q0FC8vLzs4cowDCZMmMDgwYNp0aIFFStW5JtvvuHUqVMsWLAgDb+ZpDu5c8PChfDuu+bzCRNwbtoUN62HJSIiIiKpwMXKD4+JiWHbtm0MGjTIvs3JyYmGDRuyYcOGZB1j2rRptG3blmzZsgFw5MgRzpw5Q8OGDe375MyZk+DgYDZs2EDbtm0THSM6Opro6Gj788jISABiY2OJtXhk4/bnW11HpjJ0KLZKlXB+5RWc1qyh7u7dxBUrBo8/bnVlmY7OX8dSfx1PPXYs9dex1F/HUn8dKz31NyU1WBquLly4QFxcHL6+vgm2+/r6snfv3vu+f/PmzezatYtp06bZt505c8Z+jDuPefu1O40cOZJhw4Yl2r58+XK8vLzuW0daCA0NtbqEzMXNjewjR1Lt44/JcfIkcQ0a8He3bhxr0MDqyjIlnb+Opf46nnrsWOqvY6m/jqX+OlZ66O/169eTva+l4ephTZs2jQoVKlCtWrWHOs6gQYPo27ev/XlkZCSFChWicePGeHt7P2yZDyU2NpbQ0FAaNWqEq6urpbVkRrHPPMPpli3x37yZRz/9lIqxscSPHQtublaXlino/HUs9dfx1GPHUn8dS/11LPXXsdJTf29f1ZYcloYrHx8fnJ2dOXv2bILtZ8+exc/P757vjYqKYs6cOQwfPjzB9tvvO3v2LP7+/gmOWbly5SSP5e7ujru7e6Ltrq6ulv9l3paeaslUfHzYPHAgT+3YgfPw4Th/+SXOO3fCvHnwn/NHHo7OX8dSfx1PPXYs9dex1F/HUn8dKz30NyWfb+mEFm5ubgQFBREWFmbfFh8fT1hYGNWrV7/ne+fNm0d0dDQvvfRSgu1FixbFz88vwTEjIyPZtGnTfY8pWZSTE/HvvmtOdpEzJ6xbB0FB5vTtIiIiIiLJZPlsgX379mXq1KnMmjWLPXv20L17d6KioujUqRMA7du3TzDhxW3Tpk2jZcuW5M2bN8F2m81G7969+fDDD/ntt9/YuXMn7du3p0CBArRs2TItvpJkVE8+CVu2QNmycPo01KkDU6aAYVhdmYiIiIhkAJbfc9WmTRvOnz/PkCFDOHPmDJUrV2bp0qX2CSmOHTuGk1PCDLhv3z7Wrl3L8uXLkzxm//79iYqKokuXLly+fJlatWqxdOlSPDw8HP59JIMrWRI2bYJOnWD+fOjWzQxcn30GOn9ERERE5B4sD1cAPXv2pGfPnkm+tmrVqkTbSpcujXGP0QSbzcbw4cMT3Y8lkizZs8OPP8Lo0fDOOzBtGuzcCT/9BAULWl2diIiIiKRTll8WKJIu2WwwYAAsWWIuPrx5s3kf1po1VlcmIiIiIumUwpXIvTRuDFu3QqVKcO4cNGgAkybpPiwRERERSUThSuR+ihWD9evhxRfh1i3o1Qs6dIAbN6yuTERERETSEYUrkeTw8oJvv4Xx48HZGWbPhpo1ISLC6spEREREJJ1QuBJJLpsNeveGP/6AfPngr7+galXzuYiIiIhkeQpXIilVty5s22YGq4sXISQExozRfVgiIiIiWZzClciDKFQI/vzTXA8rPh7694e2bSEqyurKRERERMQiClciD8rDw1wD6/PPwcXFXBvr8cfh4EGrKxMRERERCyhciTwMmw26d4eVK8HPD3btgsceM9fHEhEREZEsReFKJDXUqmXeh1W9Oly+DE8+CR99ZF4yKCIiIiJZgsKVSGopUMAcwera1ZzcYvBgeOYZiIy0ujIRERERSQMKVyKpyd0dvvwSpk4FNzdYsACCg2HfPqsrExEREREHU7gScYTXXoM1ayAgAPbuNe/D+vVXq6sSEREREQdSuBJxlOBg8z6s2rXh6lVo2RKGDNF9WCIiIiKZlMKViCP5+kJYGLz5pvn8gw/g6afNSS9EREREJFNRuBJxNFdXmDgRvvnGXBtr0SLzMsHdu62uTERERERSkcKVSFp5+WVYtw6KFDEXGg4OhnnzrK5KRERERFKJwpVIWqpSBbZuhQYNICoKnn8eBgyAuDirKxMRERGRh6RwJZLWfHxg6VLo1898Pno0NG0KFy9aW5eIiIiIPBSFKxEruLiYoWrOHPDygtBQqFoVtm+3ujIREREReUAKVyJWatMGNm6EYsUgIgJq1IDvvrO6KhERERF5AApXIlarUMG8D6tpU7hxA156Cfr0gdhYqysTERERkRRQuBJJD3LnhoUL4d13zecTJkCjRnDunKVliYiIiEjyKVyJpBfOzvDhh/Dzz5A9O6xeDUFBsGWL1ZWJiIiISDIoXImkN61awebNULo0nDgBtWvDjBlWVyUiIiIi96FwJZIelSkDmzbB009DdDS88gr06AExMVZXJiIiIiJ3oXAlkl7lzAm//ALDh4PNBp9/DvXrw+nTVlcmIiIiIklQuBJJz5yc4L33zMkucuaEdevM+7A2bLC6MhERERG5g8KVSEbw5JPmxBZly5ojV3XqwJQpYBhWVyYiIiIi/0/hSiSjKFnSvA/r2WfNNbC6dYPOneHmTasrExEREREUrkQyluzZ4ccf4eOPzUsGp00zR7GOH7e6MhEREZEsT+FKJKOx2WDAAFi6FPLkMadtDwoy18USEREREcsoXIlkVI0awdatUKkSnD8PDRrApEm6D0tERETEIgpXIhlZ0aKwfj20awdxcdCrF7RvD9evW12ZiIiISJajcCWS0Xl5wezZMH48ODvDt99CrVoQEWF1ZSIiIiJZisKVSGZgs0Hv3vDHH5AvH/z1F1Staj4XERERkTShcCWSmdStC9u2mcHq4kUICYExY3QfloiIiEgaULgSyWwKFYI//4ROnSA+Hvr3h7ZtISrK6spEREREMjWFK5HMyMPDXAPr88/BxcVcG+vxx+HgQasrExEREcm0FK5EMiubDbp3h5Urwc8Pdu2Cxx6DJUusrkxEREQkU1K4EsnsatUy78OqXh0uX4Ynn4SPPjIvGRQRERGRVKNwJZIVFChgjmB17WpObjF4MDzzDERGWl2ZiIiISKahcCWSVbi7w5dfwtSp4OYGCxZAcDDs22d1ZSIiIiKZgsKVSFbz2muwZg0EBMDeveZ9WL/+anVVIiIiIhmewpVIVhQcbN6H9cQTcPUqtGwJQ4boPiwRERGRh6BwJZJV+frCH39Ar17m8w8+gObNzUkvRERERCTFFK5EsjJXV5gwAWbPNtfGWrzYvExw1y6rKxMRERHJcBSuRAReegnWrYMiRcyFhh9/HObNs7oqERERkQxF4UpETFWqwNat0KABREXB88/DgAEQF2d1ZSIiIiIZgsKViPyPjw8sXQr9+pnPR4+Gpk3h4kVr6xIRERHJABSuRCQhFxczVM2ZA15eEBoKVavC9u1WVyYiIiKSrilciUjS2rSBjRuhWDGIiIAaNeC776yuSkRERCTdUrgSkburUMG8D6tpU7hxw5z4ok8fiI21ujIRERGRdEfhSkTuLXduWLgQ3n3XfD5hAjRqBOfOWVqWiIiISHqjcCUi9+fsDB9+CD//DNmzw+rVEBQEW7ZYXZmIiIhIuqFwJSLJ16oVbN4MpUvDiRNQuzbMmGF1VSIiIiLpgsKViKRMmTKwaRM8/TRER8Mrr0CPHhATY3VlIiIiIpZSuBKRlMuZE375BYYPB5sNPv8c6teH06etrkxERETEMgpXIvJgnJzgvffMyS5y5oR168z7sNavN1+Pi8O2ejUBa9ZgW70a4uKsrVdERETEwSwPV5MnTyYwMBAPDw+Cg4PZvHnzPfe/fPkyPXr0wN/fH3d3d0qVKsXixYvtr7///vvYbLYEj0ceecTRX0Mk63rySXNii3LlzJGrunWha1cIDMSlUSOqjhuHS6NGEBhoToghIiIikklZGq7mzp1L3759GTp0KOHh4VSqVImQkBDO3WWK55iYGBo1akRERATz589n3759TJ06lYCAgAT7lStXjtOnT9sfa9euTYuvI5J1lSxpLjj87LPmGlhffWVOePFfJ0+arytgiYiISCblYuWHjxs3js6dO9OpUycAvvzySxYtWsT06dMZOHBgov2nT5/Ov//+y/r163F1dQUgMDAw0X4uLi74+fk5tHYRuUP27PDDDxAaCleuJH7dMMz7s3r3hhYtzOndRURERDIRy8JVTEwM27ZtY9CgQfZtTk5ONGzYkA0bNiT5nt9++43q1avTo0cPfv31V/Lly8eLL77IgAEDcP7PP9QOHDhAgQIF8PDwoHr16owcOZLChQvftZbo6Giio6PtzyMjIwGIjY0lNjb2Yb/qQ7n9+VbXkVmpv6nLtno1LkkFq9sMA44f59bKlRh16qRdYZmUzl/HU48dS/11LPXXsdRfx0pP/U1JDZaFqwsXLhAXF4evr2+C7b6+vuzduzfJ9xw+fJgVK1bQrl07Fi9ezMGDB3n99deJjY1l6NChAAQHBzNz5kxKly7N6dOnGTZsGLVr12bXrl3kyJEjyeOOHDmSYcOGJdq+fPlyvLy8HvKbpo7Q0FCrS8jU1N/UEbBmDVWTsd/2xYs5GRXl8HqyCp2/jqceO5b661jqr2Opv46VHvp7/fr1ZO9rMwzDcGAtd3Xq1CkCAgJYv3491atXt2/v378/q1evZtOmTYneU6pUKW7evMmRI0fsI1Xjxo1jzJgxnL7LFNCXL1+mSJEijBs3jldffTXJfZIauSpUqBAXLlzA29v7Yb7mQ4uNjSU0NJRGjRrZL4WU1KP+pi7b6tXm5BX3Yfj7E//668R36AC6hPeB6fx1PPXYsdRfx1J/HUv9daz01N/IyEh8fHy4cuXKfbOBZSNXPj4+ODs7c/bs2QTbz549e9f7pfz9/XF1dU1wCWCZMmU4c+YMMTExuLm5JXpPrly5KFWqFAcPHrxrLe7u7ri7uyfa7urqavlf5m3pqZbMSP1NJfXqQcGC5uQVd/u9jc2G7fRpnN97D+dhw8zFiLt0gUaNzOndJcV0/jqeeuxY6q9jqb+Opf46Vnrob0o+37J/ybi5uREUFERYWJh9W3x8PGFhYQlGsv6rZs2aHDx4kPj4ePu2/fv34+/vn2SwArh27RqHDh3C398/db+AiCTm7AwTJ5p/ttkSvmazmY/vvoOZM6FGDbh1y5w9sEkTKFYMPvwQTp1K87JFREREUoOlvybu27cvU6dOZdasWezZs4fu3bsTFRVlnz2wffv2CSa86N69O//++y+9evVi//79LFq0iBEjRtCjRw/7Pm+//TarV68mIiKC9evX06pVK5ydnXnhhRfS/PuJZEmtW8P8+XDHEgkULGhuf+EF6NDBXHR45054803IlQuOHjUXJS5cGFq2hEWLtPCwiIiIZCiWTsXepk0bzp8/z5AhQzhz5gyVK1dm6dKl9kkujh07htN/LhMqVKgQy5Yto0+fPlSsWJGAgAB69erFgAED7PucOHGCF154gYsXL5IvXz5q1arFxo0byZcvX5p/P5Esq3VraNGCWytXsn3JEio3bYpLvXqJp18vX94c6fr4Y/jpJ3N9rD//hF9/NR+FCsGrr8Irr5h/FhEREUnHLA1XAD179qRnz55JvrZq1apE26pXr87GjRvverw5c+akVmki8jCcnTHq1OFkVBSV6tS597pWnp7w0kvmY88emDoVZs2C48fh/fdh+HBo2tS8N6tZM3Cx/EeXiIiISCK6e1xE0pcyZWDcOHNSjO+/h7p1IT7evEywRQsoUsS8fDAiwupKRURERBJQuBKR9MnDw7w/a+VK2LcP+vUDHx9zwosPPzQnwGjaFH75BdLBAoMiIiIiClcikv6VKgWjR8OJEzB3LjRsaE71vnSpeX9X4cLwzjtw+LDVlYqIiEgWpnAlIhmHuzs8/zyEhsLBgzBwIPj6wpkzMHIkFC8OjRubsxLGxFhdrYiIiGQxClcikjEVL24GquPHzZkGQ0LMdbRCQ+G558yp3wcMgAMHrK5UREREsgiFKxHJ2FxdzUsDly6FQ4fg3XfB3x/OnzcvJSxVCurXhzlzIDra6mpFREQkE1O4EpHMo2hRc7KLY8dgwQJz2nabzZwU44UXzIWN33oL9u61ulIRERHJhBSuRCTzcXExp21ftMicsn3oUDNYXbxoTvNepgzUqQPffQc3b1pdrYiIiGQSClcikrkVLmwuRBwRAQsXQvPm4OQEa9aYixYXKAC9e8Pu3RYXKiIiIhmdwpWIZA0uLvDUU/Dbb3D0KAwfbgavS5dg4kQoXx5q1YJvvoHr162uVkRERDIghSsRyXoKFoT33jPXxVqyBFq1AmdnWLcOOnQwR7PeeAP+/tvqSkVERCQDUbgSkazL2RmaNIGffzandP/oI3NSjCtX4LPPoFIlePxxmD4doqKsrlZERETSOYUrEREwp29/5x1zceLly+HZZ81LCTdtgldfNV9//XX46y+rKxUREZF0SuFKROS/nJygUSOYNw9OnIBRo8wFi69ehS++gCpV4LHHYOpUc5uIiIjI/1O4EhG5G19f6N8f9u+HsDBo08ZctHjrVujSxbw3q0sX87lhWF2tiIiIWEzhSkTkfpycoH59mDMHTp6EsWOhVCm4ds0cwXrsMQgKMke2rlyxuloRERGxiMKViEhK5MsHb70Fe/fC6tXQrh24u5v3Yr3+ujma9eqr5r1aGs0SERHJUlI1XN28eZOxY8em5iFFRNInmw2eeAK+/dYczRo/HsqUMdfImj7dnGWwUiVz1sHLl62uVkRERNJAisPV+fPn+f3331m+fDlxcXEAxMbGMnHiRAIDA/n4449TvUgRkXQtb17o3Rt274a1a6F9e/DwgJ07zfWyChSAjh3NdbQ0miUiIpJppShcrV27lpIlS/L000/TtGlTatSowT///EO5cuWYMmUK77//PsePH3dUrSIi6ZvNBjVrwqxZcOoUTJoE5cvDjRvmtlq1zOcTJ8K//1pdrYiIiKSyFIWrwYMH06xZM/7++2/69u3Lli1baNWqFSNGjOCff/6hW7dueHp6OqpWEZGMI3duc9Tq779hwwbo1Ak8PeGff8xRrgIF4KWXYM0ajWaJiIhkEikKVzt37mTw4MGUL1+e4cOHY7PZGD16NM8++6yj6hMRydhsNvP+q+nT4fRp+PxzqFwZoqPhu++gTh3zXq1PPoELF6yuVkRERB5CisLVpUuX8PHxAcDT0xMvLy/Kly/vkMJERDKdnDmhe3cID4ctW6BzZ8iWDfbtg7ffhoAAeOEFWLlSo1kiIiIZkEtK3/DPP/9w5swZAAzDYN++fURFRSXYp2LFiqlTnYhIZmSzQdWq5uOTT+CHH+Crr2DbNnMtrTlzoEQJM3x17Aj581tdsYiIiCRDisNVgwYNMP7zG9WnnnoKAJvNhmEY2Gw2+yyCIiJyHzlyQJcu5iM83FyU+Lvv4OBBGDAABg+Gli3NoNWggbmgsYiIiKRLKQpXR44ccVQdIiJSpQp88QWMGQNz55pBa9MmmDfPfBQtaoasTp3Az8/qakVEROQOKQpXRYoUcVQdIiJyW/bs8Oqr5mPHDjNkzZ4NR47AO+/AkCHQvDm2V14BXSkgIiKSbqTo+pLRo0dz48YN+/N169YRHR1tf3716lVef/311KtORCSrq1QJPvvMnGlw5kyoUQNu3YJffsGleXMadeuG00cfwcmTVlcqIiKS5aUoXA0aNIirV6/anzdt2pST//k/9OvXrzNlypTUq05ERExeXtChA6xbB7t2Qa9eGLlz43X+PM7DhkHhwtCiBSxapNEsERERi6QoXBl3TA1853MREUkD5crBhAnciohgW+/exNeqBfHx8Ntv8NRTEBgI778Px49bXamIiEiWommnREQyKk9PTtStS9yKFfDPP9C3L+TNCydOwLBhZsh66in49VfzUkIRERFxKIUrEZHMoEwZc82sEyfg+++hXj1zNGvRInMq9yJF4L33ICLC6kpFREQyrRSvc/X111+TPXt2AG7dusXMmTPx8fEBSHA/loiIWMDDA154wXzs3w9ffw0zZsCpU/Dhh/DRR9C4sbmuVvPm4OpqdcUiIiKZRorCVeHChZk6dar9uZ+fH7Nnz060j4iIpAOlSsHo0Wao+vVX+Oor+OMPWLbMfPj6mmtmvfYaFC9udbUiIiIZXorCVYQuJxERyXjc3OC558zHoUP/G806exY+/th8NGxojma1aGHuLyIiIimWonuuVqxYQdmyZYmMjEz02pUrVyhXrhx//vlnqhUnIiKprHhxGDnSnEnwp58gJARsNnNE6/nnoWBB6N8fDhywulIREZEMJ0XhasKECXTu3Blvb+9Er+XMmZOuXbsybty4VCtOREQcxNUVWreGpUvh8GEYPBj8/eH8eRgzxryksH59mDMH/rNYvIiIiNxdisLVjh07aNKkyV1fb9y4Mdu2bXvookREJA0FBsIHH8CxY7BgATz5pDmatXKlOTFGQAC89Rbs3Wt1pSIiIulaisLV2bNncb3HzFIuLi6cP3/+oYsSERELuLiY91z9/rs5ZfvQoeZlghcvwrhx5nTvTzwB334LN25YXa2IiEi6k6JwFRAQwK5du+76+t9//42/v/9DFyUiIhYrXBjefx+OHIGFC+Hpp8HJCf78E15+2RzN6tULdu+2ulIREZF0I0XhqlmzZrz33nvcvHkz0Ws3btxg6NChPPXUU6lWnIiIWMzFBZ56ypzK/ehRGD7cDF6XLsGkSVC+PNSsCbNmwfXrVlcrIiJiqRSFq8GDB/Pvv/9SqlQpRo8eza+//sqvv/7KqFGjKF26NP/++y/vvvuuo2oVERErFSwI771nToCxZAm0agXOzrB+PXTsCAUKQM+e8PffVlcqIiJiiRStc+Xr68v69evp3r07gwYNwjAMAGw2GyEhIUyePBlfX1+HFCoiIumEszM0aWI+Tp+GmTNh6lTzEsLJk81HcLC5blabNpAtm9UVi4iIpIkUjVwBFClShMWLF3PhwgU2bdrExo0buXDhAosXL6Zo0aKOqFFERNIrf38YNAgOHoTly+HZZ81LCTdtgldfNV/v3h3Cw62uVERExOFSHK5uy507N4899hjVqlUjd+7cqVmTiIhkNE5O0KgRzJsHJ07AqFFQogRcvQpffglBQVC1Knz1lblNREQkE3rgcCUiIpIkX1/o3x/27YOwMGjb1ly0eNs26NrVHM3q0gW2bIH/v7xcREQkM1C4EhERx3Bygvr14Ycf4ORJGDsWSpeGqCjzHq1q1aBKFfjiC7hyxepqRUREHprClYiIOF6+fPDWW7BnD6xeDe3agbs7bN8Or79uzjT4yiuwcaNGs0REJMNSuBIRkbRjs8ETT8C338KpUzBhApQta66RNWMGVK8OlSrBp5+aa2mJiIhkIApXIiJijTx5oFcv2LUL1q6F9u3BwwN27oQ33zRHszp0gHXrNJolIiIZgsKViIhYy2aDmjVh1ixzNOvTT6FCBbh5E775BmrVgnLlzFGuixetrlZEROSuFK5ERCT9yJ0bevaEHTtgwwbzPiwvL/NerT59ICAAXnrJvG9Lo1kiIpLOKFyJiEj6Y7PB44/DtGnmaNbnn0PlyhAdDd99B3XrQpky8MkncOGC1dWKiIgAClciIpLe5cwJ3btDeLi5NlbnzpAtm7mO1ttvm6NZbdvCihUQH291tSIikoUpXImISMZgs0HVqvDVV3D6tPm/VatCTAzMnQsNGpjraI0aBWfPWl2tiIhkQQpXIiKS8eTIYY5gbdkC27ZBt27mtoMHYeBAKFgQnnsOQkM1miUiImlG4UpERDK2KlXgiy/Me7OmTYPgYLh1C+bPh8aNoUQJGDHCHO0SERFxIIUrERHJHLJnN2cX3LjRnG2wZ0/zfq0jR+Ddd6FQIWjdGpYsgbg4q6sVEZFMSOFKREQyn4oVzfWyTp2CmTOhRg0zUP3yCzRrBsWKwQcfwMmTdz9GXBy21asJWLMG2+rVCmQiInJfloeryZMnExgYiIeHB8HBwWzevPme+1++fJkePXrg7++Pu7s7pUqVYvHixQ91TBERyaS8vKBDB1i3Dnbtgl69zLW0jh2DIUOgcGF4+mn4/XfzUsLbfv4ZAgNxadSIquPG4dKoEQQGmttFRETuwtJwNXfuXPr27cvQoUMJDw+nUqVKhISEcO7cuST3j4mJoVGjRkRERDB//nz27dvH1KlTCQgIeOBjiohIFlGuHEyYYI5WffstPPGEOdnFwoXQvDkULQpDh8KUKfDss3DiRML3nzxpblfAEhGRu7A0XI0bN47OnTvTqVMnypYty5dffomXlxfTp09Pcv/p06fz77//smDBAmrWrElgYCB16tShUqVKD3xMERHJYjw9oV07WL0a9uyBvn0hb14zTA0fbs48aBiJ33d7W+/eukRQRESS5GLVB8fExLBt2zYGDRpk3+bk5ETDhg3ZsGFDku/57bffqF69Oj169ODXX38lX758vPjiiwwYMABnZ+cHOiZAdHQ00dHR9ueRkZEAxMbGEhsb+7Bf9aHc/nyr68is1F/HUn8dS/1NBcWLw8cfw7Bh2BYswGnsWJx27Lj7/oYBx49za+VKjDp10q7OTErnsGOpv46l/jpWeupvSmqwLFxduHCBuLg4fH19E2z39fVl7969Sb7n8OHDrFixgnbt2rF48WIOHjzI66+/TmxsLEOHDn2gYwKMHDmSYcOGJdq+fPlyvLy8HuDbpb7Q0FCrS8jU1F/HUn8dS/1NJTlyENCgAVXvFa7+3/YlSzgZFZUGRWUNOocdS/11LPXXsdJDf69fv57sfS0LVw8iPj6e/Pnz89VXX+Hs7ExQUBAnT55kzJgxDB069IGPO2jQIPr27Wt/HhkZSaFChWjcuDHe3t6pUfoDi42NJTQ0lEaNGuHq6mppLZmR+utY6q9jqb+pz5YtG4wbd9/9Hj15kkqlS5sjX/LAdA47lvrrWOqvY6Wn/t6+qi05LAtXPj4+ODs7c/bs2QTbz549i5+fX5Lv8ff3x9XVFWdnZ/u2MmXKcObMGWJiYh7omADu7u64u7sn2u7q6mr5X+Zt6amWzEj9dSz117HU31RUrx4ULGhOXpHUfVf/z3nuXJznzoUGDaBLF2jZEtzc0q7OTEbnsGOpv46l/jpWeuhvSj7fsgkt3NzcCAoKIiwszL4tPj6esLAwqlevnuR7atasycGDB4mPj7dv279/P/7+/ri5uT3QMUVEROycnWHiRPPPNlvC12w289GvHzRpYv45LAzatDEDWb9+sH9/2tcsIiLphqWzBfbt25epU6cya9Ys9uzZQ/fu3YmKiqJTp04AtG/fPsHkFN27d+fff/+lV69e7N+/n0WLFjFixAh69OiR7GOKiIjcU+vWMH8+/GeZD8AMUPPnw+jRsGQJHD4MgweDvz+cPw9jx0Lp0ubo1w8/wM2b1tQvIiKWsfSeqzZt2nD+/HmGDBnCmTNnqFy5MkuXLrVPSHHs2DGcnP6X/woVKsSyZcvo06cPFStWJCAggF69ejFgwIBkH1NEROS+WreGFi24tXIl25csoXLTprjUq2eObN0WGAgffGCujbV4MXz1lRm6Vq0yH3nzQvv20LkzlClj0RcREZG0ZPmEFj179qRnz55JvrZq1apE26pXr87GjRsf+JgiIiLJ4uyMUacOJ6OiqFSnTsJg9V8uLvD00+bj2DGYPh2mTTPXzRo/3nzUrm3em/XMM+Y6WyIikilZelmgiIhIplK4MLz/PkREwO+/m4HLyQn+/BNefhkKFIBevWDXLqsrFRERB1C4EhERSW3OzvDkk/Drr+Zo1gcfmMHr8mWYNAkqVIAaNWDmTEjB+ikiIpK+KVyJiIg4UkCAOfHF4cPmPVmtWpnha8MG6NTJHM3q2ROSsXixiIikbwpXIiIiacHZ2ZzC/eef4fhxGDECihaFK1dg8mSoXBmCg837ta5ds7paERF5AApXIiIiac3fHwYNgoMHITQUnnvOnBhj82Z47TVzNKtbNwgPt7pSERFJAYUrERERqzg5QcOG8OOP5uyCo0ZBiRJw9SpMmQJBQVC1qjnN+9WrVlcrIiL3oXAlIiKSHvj6Qv/+sG8frFgBbduCmxts2wZdu5qjXZ07w5YtYBhWVysiIklQuBIREUlPnJygXj344QdzNGvsWChdGqKi4OuvoVo1qFIFPv/cvF9LRETSDYUrERGR9CpfPnjrLdizB1avhnbtwN0dtm+HHj3Me7NeecWceVCjWSIillO4EhERSe9sNnjiCfj2Wzh1CiZMgLJlzTWyZsww18yqWBE+/RQuXbK6WhGRLEvhSkREJCPJkwd69YJdu2DdOujQATw8zOdvvmmOZrVvD2vXajRLRCSNKVyJiIhkRDabOWI1c6Y5mvXpp1ChAty8CbNnQ+3aUK6cOcp18aLV1YqIZAkKVyIiIhld7tzQsyfs2AEbN5r3YXl5mfdq9ekDAQHm/VqrV2s0S0TEgRSuREREMgubDYKDYdo0czTr88+hcmWIjobvv4e6deGRR+CTT+D8eaurFRHJdBSuREREMqOcOaF7dwgPN9fG6twZsmeH/fvh7bfN0ay2bSEsDOLjra5WRCRTULgSERHJzGw2qFoVvvrKHM366ivzeWwszJ0LDRtCqVIwahScPWt1tSIiGZrClYiISFaRI4c5grVlizmi1b27ue3QIRg4EAoWhGefheXLNZolIvIAFK5ERESyokcfNe/JOn3avEcrOBhu3YKffoKQEChRAkaMMF8XEZFkUbgSERHJyrJlM2cX3LjRnG2wZ0/zfq0jR+Ddd6FQIWjVCpYsgbg4q6sVEUnXFK5ERETEVLGiuV7WqVPm+lk1a5qBasECaNYMihWDDz6AkyetrlREJF1SuBIREZGEvLygQwdYuxZ27YJevcy1tI4dgyFDoHBhePppWLjQvJRQREQAhSsRERG5l3LlYMIEc7Tq22/hiSfMyS4WLjQDVmAgDB1qBi8RkSxO4UpERETuz9MT2rWD1athzx546y3Im9cMXcOHmyHrySfNSwhjY62uVkTEEgpXIiIikjKPPAJjx5rB6ocfoF49MAxYvNic/KJIERg82JwUQ0QkC1G4EhERkQfj7g5t28KKFbB/P/TvD/nymdO3f/QRFC9uTuv+008azRKRLEHhSkRERB5eyZIwahScOAE//giNGpmjWcuXmwsTFyoEgwaZCxaLiGRSClciIiKSetzc4LnnzFB18KAZqHx94exZ+Phjc3Hihg3NABYTY3W1IiKpSuFKREREHKN4cRgxAo4fh59/hiZNwGaDsDBo0waXokUpO3OmeUmhiEgmoHAlIiIijuXqak50sWQJHD4M770HBQpgO3+ekgsW4Fq+vDkpxvffw82bVlcrIvLAFK5EREQk7QQGmlO3Hz3KrZ9+4kzVqhhOTrBqlTnVe0AA9O1rTvcuIpLBKFyJiIhI2nNxwWjenE2DB3PrwAF4/30oWBD+/RfGj4eyZaF2bZg9G27csLpaEZFkUbgSERERaxUqBEOHQkQE/P47PP00ODvD2rXQvj0UKAC9esGuXVZXKiJyTwpXIiIikj44O8OTT8Kvv8LRo/DBB+aCxJcvw6RJUKEC1KgBM2fC9etWVysikojClYiIiKQ/AQEweLC5LtaSJdC6tRm+NmyATp3M0awePWDHDqsrFRGxU7gSERGR9MvZ2ZzC/aefzCndR4yAYsXgyhX4/HOoXBmCg+Hrr+HaNaurFZEsTuFKREREMgZ/f3NR4gMHIDTUXKzY1RU2b4bOnc3Xu3WD8HCrKxWRLErhSkRERDIWJydo2BB+/BFOnIDRo6FECXPkasoUCAoyH1OmQGSk1dWKSBaicCUiIiIZV/780K8f7N8PK1ZA27bg5maOXnXrZt6b1bkzbNkChmF1tSKSySlciYiISMZns0G9evDDD3DyJHzyCZQuDVFR5v1Y1arBo4+a92lduWJ1tSKSSSlciYiISObi4wN9+8KePbB6Nbz0Eri7mzML9uhh3pv1yivmzIMazRKRVKRwJSIiIpmTzQZPPAGzZ8OpUzBhApQtCzduwIwZ5ppZFSvCp5/CpUtWVysimYDClYiIiGR+efJAr16waxesWwcdOoCHh/n8zTfNe7Pat4e1azWaJSIPTOFKREREsg6bzRyxmjkTTp+Gzz6DChXg5k1zhKt2bShXDsaPh4sXra5WRDIYhSsRERHJmnLlMu/B2rEDNm4078Py8jLv1erb1xzNatfOvG9Lo1kikgwKVyIiIpK12WwQHAzTppmjWV98AZUrQ0wMfP891K0LjzwCY8fC+fNWVysi6ZjClYiIiMht3t7m+ljh4ebaWF26QPbs5jpa/fpBQIC5llZYGMTHW12tiKQzClciIiIid7LZoGpVmDLFnGnwq6/M57GxMHcuNGwIpUrBqFFw9qzV1YpIOqFwJSIiInIvOXJA587mSFZ4OHTvbm47dAgGDoSCBeHZZ2H5co1miWRxClciIiIiyfXoo/D55+a9WdOnw+OPw61b8NNPEBICxYvDRx+Zo10ikuUoXImIiIikVLZs0KkTbNhgzjbYsyfkzAkRETB4MBQuDK1awZIlEBdndbUikkYUrkREREQeRsWK8Omn5mjVrFlQs6YZqBYsgGbNoFgxGD4cTpywulIRcTCFKxEREZHU4OUF7dvD2rWwaxf06gW5c8OxYzB0KBQpAk8/DQsXmpcSikimo3AlIiIiktrKlYMJE8zRrG+/hSeeMCe7WLjQDFiBgWbgOnbM6kpFJBUpXImIiIg4iocHtGsHq1fDnj3w1luQNy+cPGleKhgYaF46uGCBOc27iGRoClciIiIiaeGRR2DsWDNYzZkD9euDYZiTXrRqZV42+O67cOSI1ZWKyANSuBIRERFJS+7u0KYNhIXB/v3Qvz/ky2dO7z5ihDmde0iIOb27RrNEMhSFKxERERGrlCwJo0aZMwnOmweNGpmjWcuXmwsTFyxoLlR88KDVlYpIMihciYiIiFjNzc0MU8uXw6FDMGgQ+PnBuXNm+CpZEho2hB9/hOhoq6sVkbtQuBIRERFJT4oVMy8PPHYMfv4ZmjQBm828jLBNG3M0q18/85LCu4mLw7Z6NQFr1mBbvVoLGYukkXQRriZPnkxgYCAeHh4EBwezefPmu+47c+ZMbDZbgoeHh0eCfTp27JhonyZNmjj6a4iIiIikHldXc6KLJUvg8GF47z0oUAAuXDAnxihdGurWhe+/h5s3//e+n3+GwEBcGjWi6rhxuDRqZM5K+PPPVn0TkSzD8nA1d+5c+vbty9ChQwkPD6dSpUqEhIRw7ty5u77H29ub06dP2x9Hjx5NtE+TJk0S7PPDDz848muIiIiIOE5goDl1+9Gj8Ouv8NRT4ORkTvHerh0EBECfPjBxonl54YkTCd9/8qS5XQFLxKEsD1fjxo2jc+fOdOrUibJly/Lll1/i5eXF9OnT7/oem82Gn5+f/eHr65toH3d39wT75M6d25FfQ0RERMTxXFzMRYgXLoSICHj/ffMywX//NRct7t3bnBDjTre39e6tSwRFHMjFyg+PiYlh27ZtDBo0yL7NycmJhg0bsmHDhru+79q1axQpUoT4+HiqVKnCiBEjKFeuXIJ9Vq1aRf78+cmdOzf169fnww8/JG/evEkeLzo6muj/3BwaGRkJQGxsLLEWT4F6+/OtriOzUn8dS/11LPXX8dRjx1J/H5KfH7zzDgwYgG35cpw+/hine/z7CcOA48e5tXIlRp06aVdnJqXz17HSU39TUoPNMJL69UbaOHXqFAEBAaxfv57q1avbt/fv35/Vq1ezadOmRO/ZsGEDBw4coGLFily5coWxY8eyZs0adu/eTcGCBQGYM2cOXl5eFC1alEOHDvHOO++QPXt2NmzYgLOzc6Jjvv/++wwbNizR9u+//x4vL69U/MYiIiIijhGwZg1Vx427735b+/bl5BNPpEFFIpnD9evXefHFF7ly5Qre3t733DfDhas7xcbGUqZMGV544QU++OCDJPc5fPgwxYsX548//qBBgwaJXk9q5KpQoUJcuHDhvg10tNjYWEJDQ2nUqBGurq6W1pIZqb+Opf46lvrreOqxY6m/qcu2erU5ecV9xDVvTvyQIVCpUhpUlXnp/HWs9NTfyMhIfHx8khWuLL0s0MfHB2dnZ86ePZtg+9mzZ/Hz80vWMVxdXXn00Uc5eI/F9YoVK4aPjw8HDx5MMly5u7vj7u6e5LGt/su8LT3Vkhmpv46l/jqW+ut46rFjqb+ppF498/6rkyeTvu/q/zkvXIjzwoVQrRp07gxt20L27GlYaOai89ex0kN/U/L5lk5o4ebmRlBQEGFhYfZt8fHxhIWFJRjJupe4uDh27tyJv7//Xfc5ceIEFy9evOc+IiIiIhmas7M5WyCY62L9l81mPoYMgeefN6d537zZDFf+/tCtG2zblvY1i2Qyls8W2LdvX6ZOncqsWbPYs2cP3bt3Jyoqik6dOgHQvn37BBNeDB8+nOXLl3P48GHCw8N56aWXOHr0KK+99hpgTnbRr18/Nm7cSEREBGFhYbRo0YISJUoQEhJiyXcUERERSROtW8P8+ebU7P9VsKC5fdgwmDvXnKp99GgoUQKuXYMpU6BqVQgKMv/8/5N7iUjKWB6u2rRpw9ixYxkyZAiVK1dm+/btLF261D69+rFjxzh9+rR9/0uXLtG5c2fKlClDs2bNiIyMZP369ZQtWxYAZ2dn/v77b55++mlKlSrFq6++SlBQEH/++WeSl/6JiIiIZCqtW0NEBLdCQ9naty+3QkPhyBFz+23580O/frB/P6xcCS+8AG5uEB5ujmIVKACvvWaObll3e75IhmPpPVe39ezZk549eyb52qpVqxI8Hz9+POPHj7/rsTw9PVm2bFlqliciIiKSsTg7Y9Spw8moKCrVqWNeMpgUmw3q1jUfFy7AN9/A1Kmwdy9Mm2Y+KlWCLl3MxYpz5kzLbyGS4Vg+ciUiIiIi6YCPD/TtC//8A2vWwEsvgbs77NgBPXqY92Z16gQbNmg0S+QuFK5ERERE5H9sNqhdG2bPhlOnzEkyypWDGzdg5kyoUQMqVoRJk+DSJaurFUlXFK5EREREJGl58sCbb8LOnbBuHXTsCJ6esGsX9Opl3pvVvj2sXavRLBEUrkRERETkfmw2c8RqxgxzNOuzz8zRq5s3zRGu2rXN0a3x4+HiRaurFbGMwpWIiIiIJF+uXOY9WNu3w6ZN8Oqr4OUFe/aY92wVKGBOfrFqlUazJMtRuBIRERGRlLPZoFo1+PprOH0avvgCHn0UYmLg+++hXj145BEYOxbOn7e6WpE0oXAlIiIiIg/H29tcHys8HLZuNaduz57dXEerXz9zUeM2bSAsDOLjra5WxGEUrkREREQk9QQFwZQp5r1ZX30Fjz0GsbHw44/QsCGUKgWjRsHZs1ZXKpLqFK5EREREJPXlyAGdO8PmzeaIVvfu5gjXoUMwcCAULAjPPgvLl2s0SzINhSsRERERcaxHH4XPPzdHs6ZPh8cfh1u34KefICQEiheHjz4yXxfJwBSuRERERCRtZMsGnTrBhg3w99/wxhuQMydERMDgwVC4MLRqBUuWQFyc1dWKpJjClYiIiIikvQoVYNIkc7Rq1iyoWdMMVAsWQLNmUKwYDB8OJ05YXalIsilciYiIiIh1vLygfXtYuxZ274bevSF3bjh2DIYOhSJFoHlzWLjQvJRQJB1TuBIRERGR9KFsWRg/3hzN+vZbqFPHnOzi99/h6achMNAMXMeOWV2pSJIUrkREREQkffHwgHbtYNUq2LMH3noLfHzg5EnzUsHAQPPSwQULzGneRdIJhSsRERERSb8eeQTGjjXvvZozB+rXB8MwJ71o1cq8bPDdd+HIEasrFVG4EhEREZEMwN0d2rSBsDA4cAAGDID8+eH0aRgxwpzOPSTEnN5do1liEYUrEREREclYSpSAjz+G48dh3jxo1MgczVq+3FyYuGBBc6HigwetrlSyGIUrEREREcmY3NzMMLV8ORw6BO+8A35+cO4cjBoFJUtCw4bw448QHW11tZIFKFyJiIiISMZXrBh89JE5k+DPP0PTpmCzmZcRtmljjmb16wf791tdqWRiClciIiIiknm4upoTXSxebE5y8d57EBAAFy6YE2OULg1168L338PNm1ZXK5mMwpWIiIiIZE5FiphTt0dEwG+/wVNPgZMTrF5tTvUeEAB9+sA//1hdqWQSClciIiIikrm5uEDz5rBwIRw9CsOGQaFC8O+/MGEClCsHtWvD7Nlw44bV1UoGpnAlIiIiIllHwYIwZIh5yeCiRdCiBTg7w9q10L49FCgAb74Ju3ZZXalkQApXIiIiIpL1ODtDs2awYIE5CcaHH5qXEV6+DJ9+ChUqQI0aMHMmXL9ucbGSUShciYiIiEjWVqAAvPsuHD4MS5dC69bmpYQbNkCnTuDvDz16wI4dVlcq6ZzClYiIiIgImJNdhITATz+ZCxSPHGlO8R4ZCZ9/DpUrQ3AwfP01XLtmdbWSDilciYiIiIjcyc8PBg6EAwfgjz/g+efNad43b4bOnXEpXJiKX3wB4eFWVyrpiMKViIiIiMjdODlBgwYwdy6cOAFjxkDJktiuXaPosmW4Pv44BAXBlCnmCJdkaQpXIiIiIiLJkT8/vP027NvHrdBQTtSujeHmZo5edetm3rv12mvm6JZhWF2tWEDhSkREREQkJWw2jDp12PbWW9yKiIBx4+CRRyAqCqZNM+/LevRR8z6tK1esrlbSkMKViIiIiMiD8vGBPn3gn39gzRp46SVwdzdnFuzRw5xpsFMnc+ZBjWZlegpXIiIiIiIPy2aD2rVh9mw4dQomToRy5eDGDXOtrBo1oGJFmDQJLl2yulpxEIUrEREREZHUlCcPvPkm7NwJ69dDx47g6Qm7dkGvXua9We3bw9q1Gs3KZBSuREREREQcwWaD6tVhxgxzNOuzz8zRq5s3zRGu2rXN0a3x4+HiRaurlVSgcCUiIiIi4mi5cpn3YG3fDps2wauvgpcX7NkDffuao1nt2sGqVRrNysAUrkRERERE0orNBtWqwddfw+nT8OWX5syCMTHw/fdQr5458+DYsXD+vNXVSgopXImIiIiIWMHbG7p2NdfJ2roVunSB7Nlh/37o1w8CAqBNGwgLg/h4q6uVZFC4EhERERGxWlAQTJlijmZNnQqPPQaxsfDjj9CwIZQqBaNGwdmzVlcq96BwJSIiIiKSXmTPDq+9Bps3w19/weuvmyNchw7BwIFQsCA8+ywsX67RrHRI4UpEREREJD2qXBkmTzZnGpw+3Zx58NYt+OknCAmB4sXho4/M1yVdULgSEREREUnPsmWDTp3MNbP+/hveeMOcfTAiAgYPhsKFoVUrWLIE4uKsrjZLU7gSEREREckoKlSASZPM0apvvoFatcxAtWABNGsGxYrB8OFw4oTVlWZJClciIiIiIhmNpye8/DL8+Sfs3g29e0OePHDsGAwdCkWKQPPmsHCheSmhpAmFKxERERGRjKxsWRg/Hk6ehO++gzp1zMkufv8dnn4aAgPNwHXsmNWVZnoKVyIiIiIimYGHB7z4IqxaBXv3wttvg4+PGbqGDzdDVrNm5iWEsbEWF5s5KVyJiIiIiGQ2pUvDmDHmvVdz5kCDBmAY5qQXrVqZlw2++y4cOWJ1pZmKwpWIiIiISGbl7g5t2sAff8CBAzBgAOTPby5WPGKEOZ17SIg5vbtGsx6awpWIiIiISFZQogR8/DEcPw7z50PjxuZo1vLl5sLEBQuaCxUfPGh1pRmWwpWIiIiISFbi5gbPPAPLlsGhQ/DOO+DnB+fOwahRULIkNGwIP/4I0dFWV5uhKFyJiIiIiGRVxYrBRx+ZMwn+8gs0bQo2G4SFmZcTFiwI/frB/v1WV5ohKFyJiIiIiGR1rq7QsiUsXmxOcjFkCAQEwIULMHasOUFG3brw/fdw86bV1aZbClciIiIiIvI/RYrAsGEQEQG//QZPPQVOTrB6NbRrZ4auPn3gn3+srjTdUbgSEREREZHEXFygeXNYuBCOHjUDV6FC8O+/MGEClCsHtWvD7Nlw44bV1aYLClciIiIiInJvBQualwoeOQKLFpmXEDo7w9q10L49FCgAb74Ju3ZZXamlFK5ERERERCR5nJ2hWTNz8otjx+DDD83LCC9fhk8/hQoVoEYNmDkTrl+3uto0p3AlIiIiIiIpV6AAvPsuHD5sTuv+zDPmpYQbNkCnTuDvDz16wI4dVleaZhSuRERERETkwTk5mQsSz59vLlA8ciQULw6RkfD551C5MgQHw9dfw7VrVlfrUApXIiIiIiKSOvz8YOBAc12sP/6A5583p3nfvBk6dzZHs7p1g/Dwux8jLg7b6tUErFmDbfVqiItLu/ofUroIV5MnTyYwMBAPDw+Cg4PZvHnzXfedOXMmNpstwcPDwyPBPoZhMGTIEPz9/fH09KRhw4YcOHDA0V9DRERERETAHM1q0ADmzoWTJ2HMGChZ0hy5mjIFgoLMx5Qp5gjXbT//DIGBuDRqRNVx43Bp1AgCA83tGYDl4Wru3Ln07duXoUOHEh4eTqVKlQgJCeHcuXN3fY+3tzenT5+2P44ePZrg9dGjRzNp0iS+/PJLNm3aRLZs2QgJCeGmFjwTEREREUlb+fLB22/Dvn2wciW8+CK4uZmjV926mfduvfYajBoFzz4LJ04kfP/Jk+b2DBCwLA9X48aNo3PnznTq1ImyZcvy5Zdf4uXlxfTp0+/6HpvNhp+fn/3h6+trf80wDCZMmMDgwYNp0aIFFStW5JtvvuHUqVMsWLAgDb6RiIiIiIgkYrNB3brw3XdmYBo3Dh55BKKiYNo083JCw0j8vtvbevdO95cIulj54TExMWzbto1BgwbZtzk5OdGwYUM2bNhw1/ddu3aNIkWKEB8fT5UqVRgxYgTlypUD4MiRI5w5c4aGDRva98+ZMyfBwcFs2LCBtm3bJjpedHQ00dHR9ueR/z80GRsbS2xs7EN/z4dx+/OtriOzUn8dS/11LPXX8dRjx1J/HUv9dSz19yHlzAk9e0KPHtjWr8fpo49w+uOPu+9vGHD8OLdWrsSoUyft6iRlf8eWhqsLFy4QFxeXYOQJwNfXl7179yb5ntKlSzN9+nQqVqzIlStXGDt2LDVq1GD37t0ULFiQM2fO2I9x5zFvv3ankSNHMmzYsETbly9fjpeX14N8tVQXGhpqdQmZmvrrWOqvY6m/jqceO5b661jqr2Opv6kjoGJFqt4rXP2/7UuWcDIqKg0q+p/rKVivy9Jw9SCqV69O9erV7c9r1KhBmTJlmDJlCh988MEDHXPQoEH07dvX/jwyMpJChQrRuHFjvL29H7rmhxEbG0toaCiNGjXC1dXV0loyI/XXsdRfx1J/HU89diz117HUX8dSf1OXLVs28zLB+6jctCmV0njkKvK/E27ch6XhysfHB2dnZ86ePZtg+9mzZ/Hz80vWMVxdXXn00Uc5ePAggP19Z8+exd/fP8ExK1eunOQx3N3dcXd3T/LY6eU/lvRUS2ak/jqW+utY6q/jqceOpf46lvrrWOpvKqlXDwoWNO/FSuq+K5sNChbEpV49cHZO09JS8vdr6YQWbm5uBAUFERYWZt8WHx9PWFhYgtGpe4mLi2Pnzp32IFW0aFH8/PwSHDMyMpJNmzYl+5giIiIiIpKGnJ1h4kTzzzZbwtduP58wIc2DVUpZPltg3759mTp1KrNmzWLPnj10796dqKgoOnXqBED79u0TTHgxfPhwli9fzuHDhwkPD+ell17i6NGjvPbaa4A5k2Dv3r358MMP+e2339i5cyft27enQIECtGzZ0oqvKCIiIiIi99O6NcyfDwEBCbcXLGhub93amrpSwPJ7rtq0acP58+cZMmQIZ86coXLlyixdutQ+IcWxY8dwcvpfBrx06RKdO3fmzJkz5M6dm6CgINavX0/ZsmXt+/Tv35+oqCi6dOnC5cuXqVWrFkuXLk202LCIiIiIiKQjrVtDixbcWrmS7UuWULlpU0suBXxQlocrgJ49e9KzZ88kX1u1alWC5+PHj2f8+PH3PJ7NZmP48OEMHz48tUoUEREREZG04OyMUacOJ6OizMkrMkiwgnRwWaCIiIiIiEhmoHAlIiIiIiKSChSuREREREREUoHClYiIiIiISCpQuBIREREREUkFClciIiIiIiKpQOFKREREREQkFShciYiIiIiIpAKFKxERERERkVSgcCUiIiIiIpIKXKwuID0yDAOAyMhIiyuB2NhYrl+/TmRkJK6urlaXk+mov46l/jqW+ut46rFjqb+Opf46lvrrWOmpv7czwe2McC8KV0m4evUqAIUKFbK4EhERERERSQ+uXr1Kzpw577mPzUhOBMti4uPjOXXqFDly5MBms1laS2RkJIUKFeL48eN4e3tbWktmpP46lvrrWOqv46nHjqX+Opb661jqr2Olp/4ahsHVq1cpUKAATk73vqtKI1dJcHJyomDBglaXkYC3t7flJ1Zmpv46lvrrWOqv46nHjqX+Opb661jqr2Oll/7eb8TqNk1oISIiIiIikgoUrkRERERERFKBwlU65+7uztChQ3F3d7e6lExJ/XUs9dex1F/HU48dS/11LPXXsdRfx8qo/dWEFiIiIiIiIqlAI1ciIiIiIiKpQOFKREREREQkFShciYiIiIiIpAKFKxERERERkVSgcJUOTJ48mcDAQDw8PAgODmbz5s333H/evHk88sgjeHh4UKFCBRYvXpxGlWZMKenvzJkzsdlsCR4eHh5pWG3GsmbNGpo3b06BAgWw2WwsWLDgvu9ZtWoVVapUwd3dnRIlSjBz5kyH15lRpbS/q1atSnT+2mw2zpw5kzYFZzAjR47kscceI0eOHOTPn5+WLVuyb9+++75PP4OT50H6q5/ByffFF19QsWJF+wKr1atXZ8mSJfd8j87d5Etpf3XuPpyPP/4Ym81G796977lfRjiHFa4sNnfuXPr27cvQoUMJDw+nUqVKhISEcO7cuST3X79+PS+88AKvvvoqf/31Fy1btqRly5bs2rUrjSvPGFLaXzBXAj99+rT9cfTo0TSsOGOJioqiUqVKTJ48OVn7HzlyhCeffJJ69eqxfft2evfuzWuvvcayZcscXGnGlNL+3rZv374E53D+/PkdVGHGtnr1anr06MHGjRsJDQ0lNjaWxo0bExUVddf36Gdw8j1If0E/g5OrYMGCfPzxx2zbto2tW7dSv359WrRowe7du5PcX+duyqS0v6Bz90Ft2bKFKVOmULFixXvul2HOYUMsVa1aNaNHjx7253FxcUaBAgWMkSNHJrn/888/bzz55JMJtgUHBxtdu3Z1aJ0ZVUr7O2PGDCNnzpxpVF3mAhi//PLLPffp37+/Ua5cuQTb2rRpY4SEhDiwsswhOf1duXKlARiXLl1Kk5oym3PnzhmAsXr16rvuo5/BDy45/dXP4IeTO3du4+uvv07yNZ27D+9e/dW5+2CuXr1qlCxZ0ggNDTXq1Klj9OrV6677ZpRzWCNXFoqJiWHbtm00bNjQvs3JyYmGDRuyYcOGJN+zYcOGBPsDhISE3HX/rOxB+gtw7do1ihQpQqFChe77WypJGZ2/aaNy5cr4+/vTqFEj1q1bZ3U5GcaVK1cAyJMnz1330Tn84JLTX9DP4AcRFxfHnDlziIqKonr16knuo3P3wSWnv6Bz90H06NGDJ598MtG5mZSMcg4rXFnowoULxMXF4evrm2C7r6/vXe+ROHPmTIr2z8oepL+lS5dm+vTp/Prrr3z77bfEx8dTo0YNTpw4kRYlZ3p3O38jIyO5ceOGRVVlHv7+/nz55Zf89NNP/PTTTxQqVIi6desSHh5udWnpXnx8PL1796ZmzZqUL1/+rvvpZ/CDSW5/9TM4ZXbu3En27Nlxd3enW7du/PLLL5QtWzbJfXXuplxK+qtzN+XmzJlDeHg4I0eOTNb+GeUcdrG6AJH0pHr16gl+K1WjRg3KlCnDlClT+OCDDyysTOT+SpcuTenSpe3Pa9SowaFDhxg/fjyzZ8+2sLL0r0ePHuzatYu1a9daXUqmlNz+6mdwypQuXZrt27dz5coV5s+fT4cOHVi9evVdA4CkTEr6q3M3ZY4fP06vXr0IDQ3NdBN/KFxZyMfHB2dnZ86ePZtg+9mzZ/Hz80vyPX5+finaPyt7kP7eydXVlUcffZSDBw86osQs527nr7e3N56enhZVlblVq1ZNgeE+evbsye+//86aNWsoWLDgPffVz+CUS0l/76Sfwffm5uZGiRIlAAgKCmLLli1MnDiRKVOmJNpX527KpaS/d9K5e2/btm3j3LlzVKlSxb4tLi6ONWvW8NlnnxEdHY2zs3OC92SUc1iXBVrIzc2NoKAgwsLC7Nvi4+MJCwu76zW91atXT7A/QGho6D2vAc6qHqS/d4qLi2Pnzp34+/s7qswsRedv2tu+fbvO37swDIOePXvyyy+/sGLFCooWLXrf9+gcTr4H6e+d9DM4ZeLj44mOjk7yNZ27D+9e/b2Tzt17a9CgATt37mT79u32R9WqVWnXrh3bt29PFKwgA53DVs+okdXNmTPHcHd3N2bOnGn8888/RpcuXYxcuXIZZ86cMQzDMF5++WVj4MCB9v3XrVtnuLi4GGPHjjX27NljDB061HB1dTV27txp1VdI11La32HDhhnLli0zDh06ZGzbts1o27at4eHhYezevduqr5CuXb161fjrr7+Mv/76ywCMcePGGX/99Zdx9OhRwzAMY+DAgcbLL79s3//w4cOGl5eX0a9fP2PPnj3G5MmTDWdnZ2Pp0qVWfYV0LaX9HT9+vLFgwQLjwIEDxs6dO41evXoZTk5Oxh9//GHVV0jXunfvbuTMmdNYtWqVcfr0afvj+vXr9n30M/jBPUh/9TM4+QYOHGisXr3aOHLkiPH3338bAwcONGw2m7F8+XLDMHTuPqyU9lfn7sO7c7bAjHoOK1ylA59++qlRuHBhw83NzahWrZqxceNG+2t16tQxOnTokGD/H3/80ShVqpTh5uZmlCtXzli0aFEaV5yxpKS/vXv3tu/r6+trNGvWzAgPD7eg6ozh9tTfdz5u97RDhw5GnTp1Er2ncuXKhpubm1GsWDFjxowZaV53RpHS/o4aNcooXry44eHhYeTJk8eoW7eusWLFCmuKzwCS6i2Q4JzUz+AH9yD91c/g5HvllVeMIkWKGG5ubka+fPmMBg0a2P/hbxg6dx9WSvurc/fh3RmuMuo5bDMMw0i7cTIREREREZHMSfdciYiIiIiIpAKFKxERERERkVSgcCUiIiIiIpIKFK5ERERERERSgcKViIiIiIhIKlC4EhERERERSQUKVyIiIiIiIqlA4UpERERERCQVKFyJiIikMpvNxoIFC6wuQ0RE0pjClYiIZCodO3bEZrMlejRp0sTq0kREJJNzsboAERGR1NakSRNmzJiRYJu7u7tF1YiISFahkSsREcl03N3d8fPzS/DInTs3YF6y98UXX9C0aVM8PT0pVqwY8+fPT/D+nTt3Ur9+fTw9PcmbNy9dunTh2rVrCfaZPn065cqVw93dHX9/f3r27Jng9QsXLtCqVSu8vLwoWbIkv/32m2O/tIiIWE7hSkREspz33nuPZ555hh07dtCuXTvatm3Lnj17AIiKiiIkJITcuXOzZcsW5s2bxx9//JEgPH3xxRf06NGDLl26sHPnTn777TdKlCiR4DOGDRvG888/z99//02zZs1o164d//77b5p+TxERSVs2wzAMq4sQERFJLR07duTbb7/Fw8MjwfZ33nmHd955B5vNRrdu3fjiiy/srz3++ONUqVKFzz//nKlTpzJgwACOHz9OtmzZAFi8eDHNmzfn1KlT+Pr6EhAQQKdOnfjwww+TrMFmszF48GA++OADwAxs2bNnZ8mSJbr3S0QkE9M9VyIikunUq1cvQXgCyJMnj/3P1atXT/Ba9erV2b59OwB79uyhUqVK9mAFULNmTeLj49m3bx82m41Tp07RoEGDe9ZQsWJF+5+zZcuGt7c3586de9CvJCIiGYDClYiIZDrZsmVLdJleavH09EzWfq6urgme22w24uPjHVGSiIikE7rnSkREspyNGzcmel6mTBkAypQpw44dO4iKirK/vm7dOpycnChdujQ5cuQgMDCQsLCwNK1ZRETSP41ciYhIphMdHc2ZM2cSbHNxccHHxweAefPmUbVqVWrVqsV3333H5s2bmTZtGgDt2rVj6NChdOjQgffff5/z58/zxhtv8PLLL+Pr6wvA+++/T7du3cifPz9Nmzbl6tWrrFu3jjfeeCNtv6iIiKQrClciIpLpLF26FH9//wTbSpcuzd69ewFzJr85c+bw+uuv4+/vzw8//EDZsmUB8PLyYtmyZfTq1YvHHnsMLy8vnnnmGcaNG2c/VocOHbh58ybjx4/n7bffxsfHh2effTbtvqCIiKRLmi1QRESyFJvNxi+//ELLli2tLkVERDIZ3XMlIiIiIiKSChSuREREREREUoHuuRIRkSxFV8OLiIijaORKREREREQkFShciYiIiIiIpAKFKxERERERkVSgcCUiIiIiIpIKFK5ERERERERSgcKViIiIiIhIKlC4EhERERERSQUKVyIiIiIiIqng/wD/7Ex2EkQs2AAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import cv2\n",
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras import backend as K\n",
        "from google.colab import files\n",
        "\n",
        "# ----------------------------\n",
        "# Helper Functions\n",
        "# ----------------------------\n",
        "\n",
        "def load_and_preprocess_uploaded_image(file_bytes, target_size=(128, 32)):\n",
        "    \"\"\"\n",
        "    Loads an image from raw bytes, converts to grayscale,\n",
        "    resizes to target_size (interpreted as (width, height) by OpenCV),\n",
        "    normalizes to [0,1], and reshapes for model input.\n",
        "    \"\"\"\n",
        "    data = np.frombuffer(file_bytes, np.uint8)\n",
        "    img = cv2.imdecode(data, cv2.IMREAD_GRAYSCALE)\n",
        "    if img is None:\n",
        "        raise ValueError(\"Error loading image from the uploaded file.\")\n",
        "\n",
        "    # cv2.resize expects (width, height).\n",
        "    # If you want the final image to be (128,32,1) (height, width, channels)\n",
        "    # then you need to swap the target size: target_size = (32,128)\n",
        "    img = cv2.resize(img, (target_size[1], target_size[0]))\n",
        "    img = img.astype('float32') / 255.0\n",
        "    img = np.expand_dims(img, axis=-1)\n",
        "    img = np.expand_dims(img, axis=0)\n",
        "    return img\n",
        "\n",
        "def decode_predictions(predictions, blank_label=-1):\n",
        "    \"\"\"\n",
        "    Decodes raw predictions using greedy CTC decoding and removes blank tokens.\n",
        "    \"\"\"\n",
        "    time_steps = predictions.shape[1]\n",
        "    batch_size = predictions.shape[0]\n",
        "    input_length = np.full((batch_size,), time_steps, dtype=np.int32)\n",
        "\n",
        "    decoded, log_prob = K.ctc_decode(predictions, input_length=input_length, greedy=True)\n",
        "    decoded_sequences = K.get_value(decoded[0])\n",
        "\n",
        "    final_sequences = []\n",
        "    for seq in decoded_sequences:\n",
        "        # Remove blank tokens (assumed to be -1)\n",
        "        seq = seq[seq != blank_label]\n",
        "        final_sequences.append(seq.tolist())\n",
        "    return final_sequences\n",
        "\n",
        "def map_sequence_to_text(seq, index_to_char):\n",
        "    \"\"\"\n",
        "    Converts a numeric sequence into a string using the provided mapping.\n",
        "    \"\"\"\n",
        "    # Here, it's assumed that 0 is reserved for the blank token.\n",
        "    # Adjust the mapping as per your training setup.\n",
        "    return \"\".join([index_to_char.get(num, \"\") for num in seq])\n",
        "\n",
        "def recognize_uploaded_image(inference_model, index_to_char):\n",
        "    \"\"\"\n",
        "    Uploads an image, preprocesses it, runs inference, decodes predictions,\n",
        "    maps the numeric output to text, and prints the recognized text.\n",
        "    \"\"\"\n",
        "    uploaded = files.upload()\n",
        "    if not uploaded:\n",
        "        print(\"No file uploaded.\")\n",
        "        return\n",
        "\n",
        "    for filename in uploaded.keys():\n",
        "        print(f\"Processing uploaded file: {filename}\")\n",
        "        file_bytes = uploaded[filename]\n",
        "        # Load and preprocess image.\n",
        "        # Note: We swap the target_size here so that the final shape is (128,32,1)\n",
        "        img = load_and_preprocess_uploaded_image(file_bytes, target_size=(128, 32))\n",
        "        # If your model expects (128,32,1) but you got (32,128,1), you can transpose:\n",
        "        if img.shape[1:3] == (32, 128):\n",
        "            img = np.transpose(img, (0, 2, 1, 3))\n",
        "            print(\"Image transposed to match expected input shape.\")\n",
        "        print(\"Image shape:\", img.shape)\n",
        "\n",
        "        predictions = inference_model.predict(img)\n",
        "        print(\"Raw predictions shape:\", predictions.shape)\n",
        "\n",
        "        decoded_sequences = decode_predictions(predictions, blank_label=-1)\n",
        "        print(\"Decoded numeric sequence:\", decoded_sequences[0])\n",
        "\n",
        "        recognized_text = map_sequence_to_text(decoded_sequences[0], index_to_char)\n",
        "        print(\"Recognized text:\", recognized_text)\n",
        "        break  # Process only the first uploaded file\n",
        "\n",
        "# ----------------------------\n",
        "# Define Your Inference Model\n",
        "# ----------------------------\n",
        "\n",
        "# Assuming that during training, you defined 'input_img' and 'y_pred'\n",
        "# and built your model as follows:\n",
        "# inference_model = tf.keras.models.Model(inputs=input_img, outputs=y_pred)\n",
        "# Here, we assume inference_model is already defined in your environment.\n",
        "\n",
        "# ----------------------------\n",
        "# Define Your Mapping Dictionary\n",
        "# ----------------------------\n",
        "\n",
        "# This mapping must be the same as used during training.\n",
        "# For example, if your classes correspond to uppercase letters, digits, etc.,\n",
        "# define the dictionary accordingly. Here is a placeholder example:\n",
        "index_to_char = {\n",
        "    # Example mapping (you must replace this with your actual mapping):\n",
        "    1: 'A', 2: 'B', 3: 'C', 4: 'D', 5: 'E', 6: 'F', 7: 'G', 8: 'H', 9: 'I',\n",
        "    10: 'J', 11: 'K', 12: 'L', 13: 'M', 14: 'N', 15: 'O', 16: 'P', 17: 'Q',\n",
        "    18: 'R', 19: 'S', 20: 'T', 21: 'U', 22: 'V', 23: 'W', 24: 'X', 25: 'Y',\n",
        "    26: 'Z',\n",
        "    # Continue for all labels...\n",
        "    224: 'M',   # Example: label 224 maps to 'M' (if that is correct for your dataset)\n",
        "    # ...\n",
        "}\n",
        "\n",
        "# ----------------------------\n",
        "# Test the Model on an Uploaded Image\n",
        "# ----------------------------\n",
        "recognize_uploaded_image(inference_model, index_to_char)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 177
        },
        "id": "21ltoE6vOZrw",
        "outputId": "b387473e-a75b-4865-d4b2-d0c0c1d6ffc6"
      },
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-43c40969-6e14-47b1-9bbe-e9bb1e55a480\" name=\"files[]\" multiple disabled\n",
              "        style=\"border:none\" />\n",
              "     <output id=\"result-43c40969-6e14-47b1-9bbe-e9bb1e55a480\">\n",
              "      Upload widget is only available when the cell has been executed in the\n",
              "      current browser session. Please rerun this cell to enable.\n",
              "      </output>\n",
              "      <script>// Copyright 2017 Google LLC\n",
              "//\n",
              "// Licensed under the Apache License, Version 2.0 (the \"License\");\n",
              "// you may not use this file except in compliance with the License.\n",
              "// You may obtain a copy of the License at\n",
              "//\n",
              "//      http://www.apache.org/licenses/LICENSE-2.0\n",
              "//\n",
              "// Unless required by applicable law or agreed to in writing, software\n",
              "// distributed under the License is distributed on an \"AS IS\" BASIS,\n",
              "// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
              "// See the License for the specific language governing permissions and\n",
              "// limitations under the License.\n",
              "\n",
              "/**\n",
              " * @fileoverview Helpers for google.colab Python module.\n",
              " */\n",
              "(function(scope) {\n",
              "function span(text, styleAttributes = {}) {\n",
              "  const element = document.createElement('span');\n",
              "  element.textContent = text;\n",
              "  for (const key of Object.keys(styleAttributes)) {\n",
              "    element.style[key] = styleAttributes[key];\n",
              "  }\n",
              "  return element;\n",
              "}\n",
              "\n",
              "// Max number of bytes which will be uploaded at a time.\n",
              "const MAX_PAYLOAD_SIZE = 100 * 1024;\n",
              "\n",
              "function _uploadFiles(inputId, outputId) {\n",
              "  const steps = uploadFilesStep(inputId, outputId);\n",
              "  const outputElement = document.getElementById(outputId);\n",
              "  // Cache steps on the outputElement to make it available for the next call\n",
              "  // to uploadFilesContinue from Python.\n",
              "  outputElement.steps = steps;\n",
              "\n",
              "  return _uploadFilesContinue(outputId);\n",
              "}\n",
              "\n",
              "// This is roughly an async generator (not supported in the browser yet),\n",
              "// where there are multiple asynchronous steps and the Python side is going\n",
              "// to poll for completion of each step.\n",
              "// This uses a Promise to block the python side on completion of each step,\n",
              "// then passes the result of the previous step as the input to the next step.\n",
              "function _uploadFilesContinue(outputId) {\n",
              "  const outputElement = document.getElementById(outputId);\n",
              "  const steps = outputElement.steps;\n",
              "\n",
              "  const next = steps.next(outputElement.lastPromiseValue);\n",
              "  return Promise.resolve(next.value.promise).then((value) => {\n",
              "    // Cache the last promise value to make it available to the next\n",
              "    // step of the generator.\n",
              "    outputElement.lastPromiseValue = value;\n",
              "    return next.value.response;\n",
              "  });\n",
              "}\n",
              "\n",
              "/**\n",
              " * Generator function which is called between each async step of the upload\n",
              " * process.\n",
              " * @param {string} inputId Element ID of the input file picker element.\n",
              " * @param {string} outputId Element ID of the output display.\n",
              " * @return {!Iterable<!Object>} Iterable of next steps.\n",
              " */\n",
              "function* uploadFilesStep(inputId, outputId) {\n",
              "  const inputElement = document.getElementById(inputId);\n",
              "  inputElement.disabled = false;\n",
              "\n",
              "  const outputElement = document.getElementById(outputId);\n",
              "  outputElement.innerHTML = '';\n",
              "\n",
              "  const pickedPromise = new Promise((resolve) => {\n",
              "    inputElement.addEventListener('change', (e) => {\n",
              "      resolve(e.target.files);\n",
              "    });\n",
              "  });\n",
              "\n",
              "  const cancel = document.createElement('button');\n",
              "  inputElement.parentElement.appendChild(cancel);\n",
              "  cancel.textContent = 'Cancel upload';\n",
              "  const cancelPromise = new Promise((resolve) => {\n",
              "    cancel.onclick = () => {\n",
              "      resolve(null);\n",
              "    };\n",
              "  });\n",
              "\n",
              "  // Wait for the user to pick the files.\n",
              "  const files = yield {\n",
              "    promise: Promise.race([pickedPromise, cancelPromise]),\n",
              "    response: {\n",
              "      action: 'starting',\n",
              "    }\n",
              "  };\n",
              "\n",
              "  cancel.remove();\n",
              "\n",
              "  // Disable the input element since further picks are not allowed.\n",
              "  inputElement.disabled = true;\n",
              "\n",
              "  if (!files) {\n",
              "    return {\n",
              "      response: {\n",
              "        action: 'complete',\n",
              "      }\n",
              "    };\n",
              "  }\n",
              "\n",
              "  for (const file of files) {\n",
              "    const li = document.createElement('li');\n",
              "    li.append(span(file.name, {fontWeight: 'bold'}));\n",
              "    li.append(span(\n",
              "        `(${file.type || 'n/a'}) - ${file.size} bytes, ` +\n",
              "        `last modified: ${\n",
              "            file.lastModifiedDate ? file.lastModifiedDate.toLocaleDateString() :\n",
              "                                    'n/a'} - `));\n",
              "    const percent = span('0% done');\n",
              "    li.appendChild(percent);\n",
              "\n",
              "    outputElement.appendChild(li);\n",
              "\n",
              "    const fileDataPromise = new Promise((resolve) => {\n",
              "      const reader = new FileReader();\n",
              "      reader.onload = (e) => {\n",
              "        resolve(e.target.result);\n",
              "      };\n",
              "      reader.readAsArrayBuffer(file);\n",
              "    });\n",
              "    // Wait for the data to be ready.\n",
              "    let fileData = yield {\n",
              "      promise: fileDataPromise,\n",
              "      response: {\n",
              "        action: 'continue',\n",
              "      }\n",
              "    };\n",
              "\n",
              "    // Use a chunked sending to avoid message size limits. See b/62115660.\n",
              "    let position = 0;\n",
              "    do {\n",
              "      const length = Math.min(fileData.byteLength - position, MAX_PAYLOAD_SIZE);\n",
              "      const chunk = new Uint8Array(fileData, position, length);\n",
              "      position += length;\n",
              "\n",
              "      const base64 = btoa(String.fromCharCode.apply(null, chunk));\n",
              "      yield {\n",
              "        response: {\n",
              "          action: 'append',\n",
              "          file: file.name,\n",
              "          data: base64,\n",
              "        },\n",
              "      };\n",
              "\n",
              "      let percentDone = fileData.byteLength === 0 ?\n",
              "          100 :\n",
              "          Math.round((position / fileData.byteLength) * 100);\n",
              "      percent.textContent = `${percentDone}% done`;\n",
              "\n",
              "    } while (position < fileData.byteLength);\n",
              "  }\n",
              "\n",
              "  // All done.\n",
              "  yield {\n",
              "    response: {\n",
              "      action: 'complete',\n",
              "    }\n",
              "  };\n",
              "}\n",
              "\n",
              "scope.google = scope.google || {};\n",
              "scope.google.colab = scope.google.colab || {};\n",
              "scope.google.colab._files = {\n",
              "  _uploadFiles,\n",
              "  _uploadFilesContinue,\n",
              "};\n",
              "})(self);\n",
              "</script> "
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Saving test1.jpg to test1 (1).jpg\n",
            "Processing uploaded file: test1 (1).jpg\n",
            "Image shape: (1, 128, 32, 1)\n",
            "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 58ms/step\n",
            "Raw predictions shape: (1, 64, 828)\n",
            "Decoded numeric sequence: [224, 0]\n",
            "Recognized text: M\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# After training, save your inference model:\n",
        "inference_model.save(\"inference_model.h5\")\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "U-MqNIdISrMy",
        "outputId": "57a56415-309d-400e-fbd7-01a9d18dba88"
      },
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:absl:You are saving your model as an HDF5 file via `model.save()` or `keras.saving.save_model(model)`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')` or `keras.saving.save_model(model, 'my_model.keras')`. \n"
          ]
        }
      ]
    }
  ]
}