CNN_LSTM_CTC.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.6"
    },
    "colab": {
      "name": "CNN-LSTM-CTC.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "QSdC-DdnAV55",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "import cv2\n",
        "import numpy as np\n",
        "import string\n",
        "import time\n",
        "import pandas as pd\n",
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import tensorflow.keras.backend as K\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "from tensorflow.keras.utils import multi_gpu_model\n",
        "from tensorflow.keras.layers import Dense, Reshape, BatchNormalization, Input, Conv2D, MaxPool2D, Lambda, Bidirectional\n",
        "\n",
        "from tensorflow.compat.v1.keras.layers import CuDNNLSTM\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.activations import relu, sigmoid, softmax\n",
        "\n",
        "from tensorflow.keras.utils import to_categorical, Sequence\n",
        "from tensorflow.keras.callbacks import ModelCheckpoint\n",
        "from tqdm import tqdm\n",
        "from collections import Counter\n",
        "\n",
        "from PIL import Image"
      ],
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C1xn3J31ZT4y",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "841c1d93-6582-453b-8bed-306ebf33bf43"
      },
      "source": [
        "tf.__version__"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'2.2.0'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WLgalMLF6-wQ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        },
        "outputId": "5b05a3ea-3f82-414b-d562-9bcb25a89063"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iftiw2LCnSR_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "### get the file from here:\n",
        "### https://github.com/rajesh-bhat/spark-ai-summit-2020-text-extraction/blob/master/mjsynth_sample.zip"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "roZXa3gOnXBR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "66640c77-ee53-403b-97b0-fb115bd38129"
      },
      "source": [
        "!wget https://github.com/rajesh-bhat/spark-ai-summit-2020-text-extraction/blob/master/mjsynth_sample.zip"
      ],
      "execution_count": 72,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2020-06-25 17:55:09--  https://github.com/rajesh-bhat/spark-ai-summit-2020-text-extraction/blob/master/mjsynth_sample.zip\n",
            "Resolving github.com (github.com)... 140.82.113.4\n",
            "Connecting to github.com (github.com)|140.82.113.4|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: unspecified [text/html]\n",
            "Saving to: ‘mjsynth_sample.zip’\n",
            "\n",
            "\rmjsynth_sample.zip      [<=>                 ]       0  --.-KB/s               \rmjsynth_sample.zip      [ <=>                ]  73.81K  --.-KB/s    in 0.03s   \n",
            "\n",
            "2020-06-25 17:55:09 (2.50 MB/s) - ‘mjsynth_sample.zip’ saved [75582]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XxUdYbFnYCbs",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "## execute the below command\n",
        "!unzip -qq mjsynth_sample.zip"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-l6JCqxQYiDJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "final_paths = []\n",
        "final_texts = []\n",
        "\n",
        "data_folder = \"mjsynth_sample\"\n",
        "for path in os.listdir(data_folder):\n",
        "\n",
        "\n",
        "    final_paths.append(data_folder + \"/\" + path)\n",
        "    final_texts.append(path.split(\"_\")[1])\n",
        "\n",
        "\n",
        "\n"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3hsfDslWY9Te",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "27f104ab-47e3-4b03-8410-8417f2200a15"
      },
      "source": [
        "final_paths[:10], final_texts[:10]"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(['mjsynth_sample/459_loyalties_45457.jpg',\n",
              "  'mjsynth_sample/377_ratchets_62528.jpg',\n",
              "  'mjsynth_sample/122_RAVAGERS_62620.jpg',\n",
              "  'mjsynth_sample/457_inheritances_39536.jpg',\n",
              "  'mjsynth_sample/106_inactivation_38579.jpg',\n",
              "  'mjsynth_sample/312_MANNA_46540.jpg',\n",
              "  'mjsynth_sample/4_NOISIER_51842.jpg',\n",
              "  'mjsynth_sample/9_loggia_45089.jpg',\n",
              "  'mjsynth_sample/46_NUNKI_52592.jpg',\n",
              "  'mjsynth_sample/34_SUCCULENT_75831.jpg'],\n",
              " ['loyalties',\n",
              "  'ratchets',\n",
              "  'RAVAGERS',\n",
              "  'inheritances',\n",
              "  'inactivation',\n",
              "  'MANNA',\n",
              "  'NOISIER',\n",
              "  'loggia',\n",
              "  'NUNKI',\n",
              "  'SUCCULENT'])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zU_i-crjAV6P",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "e0438dc7-5e97-44a6-d2a4-606360adfc0f"
      },
      "source": [
        "### get vocab from dataset\n",
        "vocab = set(\"\".join(map(str, final_texts)))\n",
        "print(sorted(vocab))"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ELOPjdTaAV6R",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "462c19e8-a35f-4df6-ec22-1c343b6ed94c"
      },
      "source": [
        "Counter(\"\".join(map(str, final_texts)))"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Counter({'0': 38,\n",
              "         '1': 56,\n",
              "         '2': 30,\n",
              "         '3': 21,\n",
              "         '4': 26,\n",
              "         '5': 16,\n",
              "         '6': 16,\n",
              "         '7': 15,\n",
              "         '8': 17,\n",
              "         '9': 10,\n",
              "         'A': 12188,\n",
              "         'B': 3874,\n",
              "         'C': 7202,\n",
              "         'D': 6379,\n",
              "         'E': 16729,\n",
              "         'F': 2799,\n",
              "         'G': 4768,\n",
              "         'H': 3898,\n",
              "         'I': 12696,\n",
              "         'J': 518,\n",
              "         'K': 1611,\n",
              "         'L': 7997,\n",
              "         'M': 4745,\n",
              "         'N': 10222,\n",
              "         'O': 8783,\n",
              "         'P': 5178,\n",
              "         'Q': 350,\n",
              "         'R': 11125,\n",
              "         'S': 14094,\n",
              "         'T': 10070,\n",
              "         'U': 4990,\n",
              "         'V': 1787,\n",
              "         'W': 1806,\n",
              "         'X': 428,\n",
              "         'Y': 2289,\n",
              "         'Z': 696,\n",
              "         'a': 21612,\n",
              "         'b': 4718,\n",
              "         'c': 10014,\n",
              "         'd': 9727,\n",
              "         'e': 31936,\n",
              "         'f': 3274,\n",
              "         'g': 8370,\n",
              "         'h': 6096,\n",
              "         'i': 24247,\n",
              "         'j': 453,\n",
              "         'k': 2761,\n",
              "         'l': 14693,\n",
              "         'm': 6928,\n",
              "         'n': 20242,\n",
              "         'o': 17329,\n",
              "         'p': 6905,\n",
              "         'q': 439,\n",
              "         'r': 19957,\n",
              "         's': 23114,\n",
              "         't': 18281,\n",
              "         'u': 9212,\n",
              "         'v': 2718,\n",
              "         'w': 2238,\n",
              "         'x': 818,\n",
              "         'y': 4643,\n",
              "         'z': 1340})"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0HR1CgWxAV6U",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "char_list = sorted(vocab)\n",
        "\n",
        "def encode_to_labels(txt):\n",
        "    # encoding each output word into digits\n",
        "    dig_lst = []\n",
        "    \n",
        "    for index, char in enumerate(txt):\n",
        "        try:\n",
        "            dig_lst.append(char_list.index(char))\n",
        "        except:\n",
        "            print(char)\n",
        "        \n",
        "    return dig_lst"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WTZyKj1LAV6W",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "d60b8772-d0d0-436b-f349-7322f26be244"
      },
      "source": [
        "encode_to_labels(\"HELLO\")"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[17, 14, 21, 21, 24]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TEbgSpZiAV6e",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_final_paths = final_paths[ : int(len(final_paths) * 0.90)]\n",
        "train_final_texts = final_texts[ : int(len(final_texts) * 0.90)]\n",
        "\n",
        "val_final_paths = final_paths[int(len(final_paths) * 0.90) : ]\n",
        "val_final_texts = final_texts[int(len(final_texts) * 0.90) : ]"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_JUD-0JpAV6g",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8d09919b-310e-46fa-d078-b768c6c8d887"
      },
      "source": [
        "len(train_final_paths), len(val_final_paths)"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(47427, 5270)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-7PMuKJ0AV6j",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "max_label_len = max([len(str(text)) for text in final_texts])"
      ],
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bsI10m1p9Ufj",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "c9152211-3b05-4a10-cf2e-12ba5b340025"
      },
      "source": [
        "max_label_len"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "23"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cQoif0EDAV6l",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class My_Generator(Sequence):\n",
        "    \n",
        "    def __init__(self, image_filenames, labels, batch_size):\n",
        "        self.image_filenames, self.labels = image_filenames, labels\n",
        "        self.batch_size = batch_size\n",
        "\n",
        "    def __len__(self):\n",
        "        return int(np.ceil(len(self.image_filenames) / float(self.batch_size)))\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        \n",
        "        batch_paths = self.image_filenames[idx * self.batch_size:(idx + 1) * self.batch_size]\n",
        "        batch_texts = self.labels[idx * self.batch_size:(idx + 1) * self.batch_size]\n",
        "        \n",
        "        images = []\n",
        "        training_txt = []\n",
        "        train_label_length = []\n",
        "        train_input_length = []\n",
        "\n",
        "        for im_path, text in zip(batch_paths, batch_texts):\n",
        "            \n",
        "            try:\n",
        "                text = str(text).strip()\n",
        "                img = cv2.cvtColor(cv2.imread(im_path), cv2.COLOR_BGR2GRAY)   \n",
        "\n",
        "                ### actually returns h, w\n",
        "                h, w = img.shape\n",
        "\n",
        "                ### if height less than 32\n",
        "                if h < 32:\n",
        "                    add_zeros = np.ones((32-h, w)) * 255\n",
        "                    img = np.concatenate((img, add_zeros))\n",
        "                    h = 32\n",
        "\n",
        "                ## if width less than 128\n",
        "                if w < 128:\n",
        "                    add_zeros = np.ones((h, 128-w)) * 255\n",
        "                    img = np.concatenate((img, add_zeros), axis=1)\n",
        "                    w = 128\n",
        "\n",
        "                ### if width is greater than 128 or height greater than 32\n",
        "                if w > 128 or h > 32:\n",
        "                    img = cv2.resize(img, (128, 32))\n",
        "\n",
        "                img = np.expand_dims(img , axis = 2)\n",
        "\n",
        "                # Normalize each image\n",
        "                img = img / 255.\n",
        "\n",
        "                images.append(img)\n",
        "                training_txt.append(encode_to_labels(text))\n",
        "                train_label_length.append(len(text))\n",
        "                train_input_length.append(31)\n",
        "            except:\n",
        "                \n",
        "                pass\n",
        "\n",
        "        return [np.array(images), \n",
        "               pad_sequences(training_txt, maxlen=max_label_len, padding='post', value=len(char_list)), \n",
        "               np.array(train_input_length), \n",
        "               np.array(train_label_length)], np.zeros(len(images))"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EV4D0xFtAV6o",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "91486932-c95c-4a29-8dc5-3beefbe4e10e"
      },
      "source": [
        "train_final_paths[:1], train_final_texts[:1]"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(['mjsynth_sample/459_loyalties_45457.jpg'], ['loyalties'])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 35
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r3BduQiCZbei",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "4525b5a2-0a8a-4eba-8321-c45fba76f5e7"
      },
      "source": [
        "!nvidia-smi"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Thu Jun 25 16:57:14 2020       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 450.36.06    Driver Version: 418.67       CUDA Version: 10.1     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla K80           Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   72C    P0    72W / 149W |   4665MiB / 11441MiB |      0%      Default |\n",
            "|                               |                      |                 ERR! |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MsVGYWb0AV6s",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "batch_size = 256\n",
        "train_generator = My_Generator(train_final_paths, train_final_texts, batch_size)\n",
        "val_generator = My_Generator(val_final_paths, val_final_texts, batch_size)"
      ],
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zeN-LJBoAV6u",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# input with shape of height=32 and width=128 \n",
        "inputs = Input(shape=(32, 128, 1))\n",
        "\n",
        "conv_1 = Conv2D(16, (3,3), activation = 'relu', padding='same')(inputs)\n",
        "pool_1 = MaxPool2D(pool_size=(2, 2), strides=2)(conv_1)\n",
        " \n",
        "conv_2 = Conv2D(32, (3,3), activation = 'relu', padding='same')(pool_1)\n",
        "pool_2 = MaxPool2D(pool_size=(2, 2), strides=2)(conv_2)\n",
        "\n",
        "conv_3 = Conv2D(64, (3,3), activation = 'relu', padding='same')(pool_2)\n",
        "conv_4 = Conv2D(64, (3,3), activation = 'relu', padding='same')(conv_3)\n",
        "pool_4 = MaxPool2D(pool_size=(2, 1))(conv_4)\n",
        " \n",
        "conv_5 = Conv2D(64, (3,3), activation = 'relu', padding='same')(pool_4)\n",
        "# Batch normalization layer\n",
        "batch_norm_5 = BatchNormalization()(conv_5)\n",
        " \n",
        "conv_6 = Conv2D(64, (3,3), activation = 'relu', padding='same')(batch_norm_5)\n",
        "batch_norm_6 = BatchNormalization()(conv_6)\n",
        "pool_6 = MaxPool2D(pool_size=(2, 1))(batch_norm_6)\n",
        " \n",
        "conv_7 = Conv2D(64, (2,2), activation = 'relu')(pool_6)\n",
        " \n",
        "squeezed = Lambda(lambda x: K.squeeze(x, 1))(conv_7)\n",
        " \n",
        "# bidirectional LSTM layers with units=128\n",
        "blstm_1 = Bidirectional(CuDNNLSTM(128, return_sequences=True))(squeezed)\n",
        "blstm_2 = Bidirectional(CuDNNLSTM(128, return_sequences=True))(blstm_1)\n",
        "\n",
        "outputs = Dense(len(char_list) + 1, activation = 'softmax')(blstm_2)\n",
        "\n",
        "# model to be used at test time\n",
        "act_model = Model(inputs, outputs)"
      ],
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vp6vRoV4AV6w",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 765
        },
        "outputId": "42f465a5-26a8-461c-df27-1ecd5dc4acc7"
      },
      "source": [
        "act_model.summary()"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_3\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_3 (InputLayer)         [(None, 32, 128, 1)]      0         \n",
            "_________________________________________________________________\n",
            "conv2d_14 (Conv2D)           (None, 32, 128, 16)       160       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_8 (MaxPooling2 (None, 16, 64, 16)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_15 (Conv2D)           (None, 16, 64, 32)        4640      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_9 (MaxPooling2 (None, 8, 32, 32)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_16 (Conv2D)           (None, 8, 32, 64)         18496     \n",
            "_________________________________________________________________\n",
            "conv2d_17 (Conv2D)           (None, 8, 32, 64)         36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_10 (MaxPooling (None, 4, 32, 64)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_18 (Conv2D)           (None, 4, 32, 64)         36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 4, 32, 64)         256       \n",
            "_________________________________________________________________\n",
            "conv2d_19 (Conv2D)           (None, 4, 32, 64)         36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 4, 32, 64)         256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_11 (MaxPooling (None, 2, 32, 64)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_20 (Conv2D)           (None, 1, 31, 64)         16448     \n",
            "_________________________________________________________________\n",
            "lambda_2 (Lambda)            (None, 31, 64)            0         \n",
            "_________________________________________________________________\n",
            "bidirectional_3 (Bidirection (None, 31, 256)           198656    \n",
            "_________________________________________________________________\n",
            "bidirectional_4 (Bidirection (None, 31, 256)           395264    \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 31, 63)            16191     \n",
            "=================================================================\n",
            "Total params: 761,151\n",
            "Trainable params: 760,895\n",
            "Non-trainable params: 256\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "08eV8H54AV6y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labels = Input(name='the_labels', shape=[max_label_len], dtype='float32')\n",
        "input_length = Input(name='input_length', shape=[1], dtype='int64')\n",
        "label_length = Input(name='label_length', shape=[1], dtype='int64')\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",
        "loss_out = Lambda(ctc_lambda_func, \n",
        "                  output_shape=(1,), \n",
        "                  name='ctc')([outputs, labels, input_length, label_length])\n",
        "\n",
        "#model to be used at training time\n",
        "model = Model(inputs=[inputs, labels, input_length, label_length], outputs=loss_out)"
      ],
      "execution_count": 40,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RiGnMMKuAV60",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "file_path = \"C_LSTM_best.hdf5\"\n",
        "\n",
        "model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer = 'adam')\n",
        "\n",
        "\n",
        "checkpoint = ModelCheckpoint(filepath=file_path, \n",
        "                             monitor='val_loss', \n",
        "                             verbose=1, \n",
        "                             save_best_only=True, \n",
        "                             mode='min')\n",
        "\n",
        "callbacks_list = [checkpoint]"
      ],
      "execution_count": 41,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "P-gsqkzoAV62",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "b5b6e8a8-eabc-4e0a-ea54-a3cf299f5ecd"
      },
      "source": [
        "epochs = 20\n",
        "\n",
        "history = model.fit(train_generator, \n",
        "                    epochs = epochs,\n",
        "                    steps_per_epoch = len(train_final_paths) // batch_size,\n",
        "                    validation_data=val_generator,\n",
        "                    validation_steps = len(val_final_paths) // batch_size,\n",
        "                    verbose = 1,\n",
        "                    callbacks = callbacks_list,\n",
        "                    shuffle=True)"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 12.0445\n",
            "Epoch 00001: val_loss improved from 22.89446 to 10.15159, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 333ms/step - loss: 12.0445 - val_loss: 10.1516\n",
            "Epoch 2/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 8.2928\n",
            "Epoch 00002: val_loss improved from 10.15159 to 8.07728, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 334ms/step - loss: 8.2928 - val_loss: 8.0773\n",
            "Epoch 3/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 6.7348\n",
            "Epoch 00003: val_loss improved from 8.07728 to 7.10487, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 335ms/step - loss: 6.7348 - val_loss: 7.1049\n",
            "Epoch 4/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 5.9091\n",
            "Epoch 00004: val_loss improved from 7.10487 to 6.31429, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 333ms/step - loss: 5.9091 - val_loss: 6.3143\n",
            "Epoch 5/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 5.3134\n",
            "Epoch 00005: val_loss improved from 6.31429 to 5.84022, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 333ms/step - loss: 5.3134 - val_loss: 5.8402\n",
            "Epoch 6/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 4.8348\n",
            "Epoch 00006: val_loss improved from 5.84022 to 5.83548, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 333ms/step - loss: 4.8348 - val_loss: 5.8355\n",
            "Epoch 7/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 4.4866\n",
            "Epoch 00007: val_loss improved from 5.83548 to 5.49891, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 63s 340ms/step - loss: 4.4866 - val_loss: 5.4989\n",
            "Epoch 8/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 4.1722\n",
            "Epoch 00008: val_loss improved from 5.49891 to 5.36147, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 63s 341ms/step - loss: 4.1722 - val_loss: 5.3615\n",
            "Epoch 9/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 3.8920\n",
            "Epoch 00009: val_loss did not improve from 5.36147\n",
            "185/185 [==============================] - 63s 341ms/step - loss: 3.8920 - val_loss: 5.4437\n",
            "Epoch 10/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 3.6984\n",
            "Epoch 00010: val_loss improved from 5.36147 to 5.32026, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 63s 339ms/step - loss: 3.6984 - val_loss: 5.3203\n",
            "Epoch 11/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 3.4805\n",
            "Epoch 00011: val_loss improved from 5.32026 to 5.08213, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 337ms/step - loss: 3.4805 - val_loss: 5.0821\n",
            "Epoch 12/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 3.2759\n",
            "Epoch 00012: val_loss improved from 5.08213 to 4.90840, saving model to C_LSTM_best.hdf5\n",
            "185/185 [==============================] - 62s 336ms/step - loss: 3.2759 - val_loss: 4.9084\n",
            "Epoch 13/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 3.0908\n",
            "Epoch 00013: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 62s 338ms/step - loss: 3.0908 - val_loss: 5.1329\n",
            "Epoch 14/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.9666\n",
            "Epoch 00014: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 62s 335ms/step - loss: 2.9666 - val_loss: 5.0559\n",
            "Epoch 15/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.8567\n",
            "Epoch 00015: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 63s 339ms/step - loss: 2.8567 - val_loss: 5.0874\n",
            "Epoch 16/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.5766\n",
            "Epoch 00016: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 63s 338ms/step - loss: 2.5766 - val_loss: 5.0566\n",
            "Epoch 17/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.5282\n",
            "Epoch 00017: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 62s 337ms/step - loss: 2.5282 - val_loss: 4.9398\n",
            "Epoch 18/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.4768\n",
            "Epoch 00018: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 63s 340ms/step - loss: 2.4768 - val_loss: 5.0134\n",
            "Epoch 19/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.2867\n",
            "Epoch 00019: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 62s 337ms/step - loss: 2.2867 - val_loss: 5.2031\n",
            "Epoch 20/20\n",
            "185/185 [==============================] - ETA: 0s - loss: 2.4280\n",
            "Epoch 00020: val_loss did not improve from 4.90840\n",
            "185/185 [==============================] - 62s 338ms/step - loss: 2.4280 - val_loss: 5.1024\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Luf5WtwAV64",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "13eaa263-3d5e-45df-e773-028430e0052e"
      },
      "source": [
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QhmP1lgvAV66",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def pre_process_image(path):\n",
        "    \n",
        "    img = cv2.cvtColor(cv2.imread(path), cv2.COLOR_BGR2GRAY)\n",
        "    ### actually returns h, w\n",
        "    h, w = img.shape\n",
        "\n",
        "    ### if height less than 32\n",
        "    if h < 32:\n",
        "        add_zeros = np.ones((32-h, w)) * 255\n",
        "        img = np.concatenate((img, add_zeros))\n",
        "        h = 32\n",
        "\n",
        "    ## if width less than 128\n",
        "    if w < 128:\n",
        "        add_zeros = np.ones((h, 128-w)) * 255\n",
        "        img = np.concatenate((img, add_zeros), axis=1)\n",
        "        w = 128\n",
        "\n",
        "    ### if width is greater than 128 or height greater than 32\n",
        "    if w > 128 or h > 32:\n",
        "        img = cv2.resize(img, (128, 32))\n",
        "\n",
        "    \n",
        "    img = np.expand_dims(img , axis = 2)\n",
        "\n",
        "    # Normalize each image\n",
        "    img = img / 255.\n",
        "    \n",
        "    return img"
      ],
      "execution_count": 45,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a2f72vfOaHAF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "act_model.load_weights('C_LSTM_best.hdf5')"
      ],
      "execution_count": 46,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kR2UGf7gAV68",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def predict_output(img):\n",
        "    # predict outputs on validation images\n",
        "    prediction = act_model.predict(np.array([img]))\n",
        "    ## shape (batch_size, num_timesteps, vocab_size)\n",
        "\n",
        "    # use CTC decoder\n",
        "    out = K.get_value(K.ctc_decode(prediction, \n",
        "                                   input_length=np.ones(prediction.shape[0]) * prediction.shape[1],\n",
        "                                   greedy=True)[0][0])\n",
        "\n",
        "    ## get the final text    \n",
        "    for x in out:\n",
        "\n",
        "        print(\"predicted text = \", end = '')\n",
        "\n",
        "        for p in x:\n",
        "\n",
        "            if int(p) != -1:\n",
        "                print(char_list[int(p)], end = '')\n",
        "        \n",
        "        print('\\n')"
      ],
      "execution_count": 68,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "du00BfeBg7Nw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "731ddfbb-151c-4a81-ac16-8805a15c970b"
      },
      "source": [
        "from IPython.display import display\n",
        "from random import randrange\n",
        "\n",
        "for i in range(10):\n",
        "    index = randrange(len(val_final_paths))\n",
        "    im = Image.open(val_final_paths[index])\n",
        "    display(im)\n",
        "\n",
        "    print(\"\\n\")\n",
        "\n",
        "    test_img = pre_process_image(val_final_paths[index])\n",
        "    predict_output(test_img)\n"
      ],
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=99x31 at 0x7F363C3E2438>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = gregarious\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=191x31 at 0x7F363E7A2BE0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = Minnesotan\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=140x31 at 0x7F363C3E2438>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = COLCBIEN\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=120x31 at 0x7F363E6C3128>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = GUARANTES\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=155x31 at 0x7F363E6C39B0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = pigeonhole\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAADoAAAAfCAIAAAA5uEn6AAAKzUlEQVR4nJVYW29TRxedmTPnfrEd4lwcJyECQitCoVABStXnPvSlv7R/ACrUqKqAthJVVJomNMax4yQkto+Pz33O5XtYyWlECdI3D5ZvZ2bvPWuvtWbo5uYmuTTKi8EYIx8beZ4zxmRZzrIsTVNZliVJSpKEc04pLYqiKApKKWOMMUYpLcuyKIqyLAkhlFJKKVb56OSEEFVV0zQtioJzTgjJsowxpihKlmWEEH7VY1cNWZaFEEmSKIpiGEaWZXmeq6palqUkSYyx4tIghEiShBCrcFGO6ssPxnQ6xVN4vCgKLKdpGiGEV3lffuaqufCToihpmqZpqigKpVQIURQFdgOlRUWreS6/kosNvGoJwzAwT57nqI6qqoQQfPxIdbHeVeHGcew4DiHE931KKZIWQuAVKAIMqiJVs32iCtUA2FBUcrEh1SQciVbpXn5/1UiSBHmTC2yhANXuV5DI8xy1xMxIo0LwRwf6gTGGQiABIQSWY+STwP/vcBwniiKAKc9z3/fjOI7jGDCQJEmSJKRBCFEUBcXGs1XoV/UxIQRR5nleJVx9SQhh/43109kzxsqy5JwDoNeuXVtYWMjzHPNi49I05ZwvLi7evXtXVVWwx+X5P1GOsizRDKqqqqoKVGDDKaX/NzN4nifLsmVZQRBwzu/evTszMxMEARBCKQUALMu6devWvXv3dnd3q2a4/HrVoJQahtFsNtvtNiGk3++fnp7+i13DMFzXBeAopZxz7CPqwRjzfb/RaIAL0zTVNC0MQyGEoii1Wu3BgwdHR0fojzzPKaUobVmWiqL0+/0oioDpKIosywrDsFarxXHMGOOcx3FMKdV1PUmSNE0ty1pbW7t582a9Xg+CoCiKJEl6vR72kxDCwKCqqoIgAUFZlit6Z4wFQeD7vmEYhmEURaEoCko4Pz9fFMXe3h6oMUmSoigkSRJCRFE0Ho9fvXoFQSGEmKap6zpjLI7jJEkYY1mWIQjP89BhWZatrKy0Wq2zszPXdfEfXdcVRUHteZZliqIAHFgJRUJCVTsvLS3du3dvMBjs7u5CeMqydBwnCII3b95gUiEEY8xxnOl0mue5bdtRFJ2dnRVFIctynueQgKWlpVarNTMz8/r168FgAHDjWc65ZVlxHL948cL3/Var1e/3syyr2omjSVFjZCPLsm3bKysrCwsLhmF0Op1Op+M4zvXr14UQOzs7qBYqWpalYRj1er3Vah0fH+/v76dpyhhL01SSpNXV1Xfv3pELccqyrN1ub2xs3Lhxw/O8drs9HA49z9N1HX8AsYxGoyzLwjDsdru+79u2XUGUV8CXJAkAXV1dBYBGo5Esyzdu3JibmwOzHh4eIjdN0zjnZ2dnd+7cefTokWVZrVbLdd0sy7rdrq7rsiyHYYjMp9MpdqPRaHzxxRczMzNnZ2fj8fj69euTyeT169eQcUVR1tbWiqKIoujLL798//59p9NBOwkhUF2Gd2DmJEksy7pz585nn312cHDw7NmzP/74oyzLhYUFSunOzs7BwQFgJEkS5/zw8HA8Hi8vLzuO888//ziO880336ysrJALPUMQFfUKIVBIlLYsyziOLctqNBqSJNXr9YcPH6JHFxcXbdsGORZFgdISQhg4Ad/Kstxut5vNZrfbff78+fHx8fLysqZpb968efr06dbWFoiGc47OCIKg3+/HcTwajZ49e/bzzz9bltVut33fH4/H0+m0LEtVVRlj6OPpdBoEQZZlpmlOJpNXr17t7e0RQtI0nU6nnud1Op00TRuNxnA4HAwGaGhFUXRdPw8XZiqOYyGEYRhLS0tFUWxtbSVJYhjG7OxsHMee50FXKaVxHBNC4jjmnKuqGgQBOl1V1e3tbdd1m82mZVnYK8/z6vU6mB8ODm0ahuEPP/yws7NjWZamaePxGFz09OnT/f19IcTu7m6325UkKcuyLMuwKIGq+b6PGgghWq3WZDIJw5Axtr6+jq08PDx0XbfRaIAm0f6TyUQIcXp6WhGcruu9Xg9JUkr39/cbjYbjOJxzzvl0OuWcy7Icx/HW1pYsy2AeCAd4Sb4YgBMwnSSJruvnLhwYQEumaep5HuccAP3666+Lonj+/HkYhrZtg3oJIUEQUEpVVRVC+L4/nU5N02SMhWE4mUwopfPz85qmUUoPDw8RBGMMz2ZZpmlaHMdRFEEm0jR1HCdJksFgANhUfF/JNTaWUnquZJAlSqnv+/h47dq1KIpOTk6CIJBlWdO0siyDIEDx1tbWms2moihhGPZ6PcZYq9UihAyHwzRNa7VaURSmaQohHMep1+u+70OowzCEWDiOc//+/bm5udFoxBibnZ0FVABTGOiPhAuxjaIIbRHHsSRJENjJZAJid1337OwMyiSE2NjY2NjYsCxL13VK6cHBAdRIkqQoihhjpmkCNkII27YfPnzYbrd1Xeec+75fr9c3Nze//fbbJ0+etNttxlgUReiqJElqtRqlFFRQGfl/LQ7qWin+ZDIBav/8889Hjx4tLCykaeq67unp6enpKaXU87y1tTVd18fjMRTR87w8z+fm5mRZhkHJ81ySJM/zut2uYRjr6+uzs7Oe52VZ9vbt2/n5+eXlZUrpX3/91ev1arUaKkIIqdVqtVoNrVmZocsWlANMKFtZlsfHx77vb2xsbG9v//77759//vmDBw8mk8n29vZ0On3//v3Kykqj0ej3+9h30zTTNB0Oh+vr6zMzM+CZyWRiGEaSJK7r/vbbb3Ec7+/vj8djwzBOTk5evnwJz9TtdqMoMk2TEAK7iJbwPA96iaJeNsfnx1fOeRRFiqK4rru/v3///v3Nzc0XL150u93bt2+naXp0dGSapm3bX331VRAEf//9N6bDnuzt7RmGYdt2q9XKsgymEVyb5/nbt29Ho5GqqpqmMcZc1z06OgI2LMvC4RSNHsdxnudhGMZxXJ3GEeF5uLAXhJA8z03THA6HnU6n2Wyurq6apvnLL7+8fPmy2WxCkB4/ftxut9+9e7e7uyvLMlpYkqRutzuZTGq12vLyMmOs3+8nSQL7Icuy7/uO45imOR6PhRCNRgMeAHKlaZqmadPpVFXVmzdvpmkaRRFc6OVw8Z6naYoa6LoOAg6C4Keffvruu+8WFxe///77wWCQpunc3Jxt20mSbG9vdzodyCM4H5mMRqPFxcWTkxN0ehAEmqYZhgFBKYoCvFt5XGw0+A7MaNu2JEm2bf/444+UUiifJEmguXPz/vjx48u3FRVcGo3GrVu30O+4ATk5Oen1ep7nua4LlwOHWpYltkjTNFjQiuQBsKtODQASbGCWZYZhOI5z+/btX3/9NQgCx3HCMETcIC5CCH3y5El1W1FdDlBKIQ1w4gC3EALPVxYHEAIesDYWrng6SRJI1H8HTCZSBeXBYVuWNRwOESUObdDn83uGDw6SlZ+0bRs7iDigfLqu53kO8YSRr+wLjItlWUmS4GwMdbzqZFZ1D5oSARVFMR6PofxYKMsyIQRKQz5x6QQ8cc4Nw4CdiKIoyzIY37IsYfKrWwVUC3YUDgTH7qvOvRXqEGh1nwK1yvMcNxiVtToP94NLp+raotFowP9DYCD6kiSFYYiV4ARQBuhFdS0ApayiuQoMFQIvgxDHUtwE4D+VxSWXTxPk0nUTzAMWA1tV+Mb+VnmDNenFlSP8BpKE/6qc9Qfjg+bGx2qXCCGIG2tVQf4P7eokQO9J6Y0AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=58x31 at 0x7F363E61CDD8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = chler\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=171x31 at 0x7F363E6C3470>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = rhizomes\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=106x31 at 0x7F363E7BB128>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = citronella\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAFYAAAAfCAIAAADyY60EAAAOh0lEQVR4nGVZ2W7jxhLl0s1902LZ8jIeZxJn8jAIAgQB8rn5nwCB85LAE4w3SdZKihTJJtlk8z6ccUfI7QfDlqnu6qpTdU4V1d9++40Qomla13VN0wghdF3HJ4qi9EdLURRFUQghVVXhGc65pmmEkLZt8YymaX3fE0IuLi6m06mu69gZz1BKy7K8v79njAkhsH/XdUIIHKcoiqZpbdt2XYeddV1XVbXv+8vLyyAIPM/TdR3fVVVVURQhxGw2W6/XQghVVeu67vvesqymaTRNy7Ls9vb23bt3YRiuVivXdYUQXddRSjVNi+P4+fmZaJqGvTRNMwxDURRcr6oqHIP/4oZCCEII3KFpmq7rsBt3kG7CV46/iD/1t4V7apqG0+UXcXrf97quU0r7vm+aRlEUSmkYhpZlwVOwCu4jhAgh6rrGY5qmCSHatoVtvu+XZSmEKIrCMAwZJFVVcTrBkqGAd9u2bZqGEILnpLOFEH3fG4bRNA2MwHbHGJH3kV74jyNwFo6QTpSbY0/EU1VVWOW6bhRFiD/CgADg97ZtOefwFFCDI2C5bdtFUTRN03Wd53lt24q3hSMopWQymcAmeQ2YW9e1erRklBzHadu2bVv58H/CfvwwbJVObN8W59y2bfkwdkNMHMchhHRddzgc+r4PgmAymURRJF0pXQBYJUmCtJKQlHvqus45b9s2y7LRaKTretM0MhKKopimGQQBuby8lJti4Ql5ExlwnLpcLquqgn1wJ4B9jOf/+EL+AhyapgkTu66TB6mqahiG4zgfP34MgqCqqufn5zzPR6PRcDhUVZVzTimV8ZDJu16vq6oCnOFoZBOAwBhzHCeO45OTk6ZpOOc4He6zbVtVVYKsk9kFQAJLxykqT/U8D1l6nAIywf4/EbAPfKRpmmVZuFKWZW3bwhTY7bpuGIbT6dT3fUVRDMMoiiIIAhTRsixlscDmTdNUVbXf74UQlFKZJtjTMAxYpev64XCoqgqfw8Jj3JGmaSilhmHI3JMGHfsF6JVlDycd58J/PCXXcfWGQb7v27Y9mUzgAkrpcc5TSvHkcDgcDAZwt2maqOT4CvglSZLNZtN13bGdMiM0TUPMkQtJkkRRZFkWUh6IQGAI4Od5nmmaMBTAbpoGvyiKwjmvqoox1jRNnufAMIhTptYxI/zHFzARd8D+lmXBAiy4uK7rqqpQtLqugz0Amuu6eZ6XZck5B5qapkmSZD6fe54n8SiJFr5gjA2HwzzPCSF5nnueF4ZhWZbAAoxRVVVTVXU8Htu2DT5X3pgZ5QRMSykNgmAwGDiOY5pmVVWGYVBKASfwMOLTdZ3ruthKEgcy3zRNYB73Pz8/n0wmIPC2bW3bppQyxoAUZCx2gHOBFNu2fd/P8/z+/n69XodhiA1VVQXUQZzIPtd1q6pC3U3TFDXV87wgCOBERIgAJ7Zt4/74qaqq4zh933POLcuSgPd9H84ryxIkDI5kjIF167pmjCFWpmnatp3nuUQHLMM1LMuyLMs0TUJIURR4BkUb3MY5z/McScoYy/Mcsd3v93EcV1XVdR3ES1VVQgjTNCmlVVU1TYM4wXIQYRAECANOMQyjruuvQm61WjHGoig6Pz9HJPu+f3x8hF/7vo+iaDgcGoYBVQOwKUe0BGWBpAW85ZOo1QCLqqooKKZpWpaFf8Ev8g4AF7IPZAZQ1HXtOA5jLI7jxWKRJIksgcgjCD4kOXBhmqbv+8ipuq6Hw2Ecx9PpVFEUSqnjOMqbDCObzeb5+dmyrJ9//tmyLMdx1uv177//Dl/CoMFg8O23304mk6qq6rrWdd1xHFVVZXLats0YkwgCyyA+yBd8IvWZ9B04AoiTVzqGjPqmKe/u7oqiyPMcZSIMQyRmmqac8/1+X5YlwO+67ng8Hg6Hvu/Xdc05Z4xdXl7KbXVdd10Xt+v7nnDOy7JEYoAk6roGzgE8EJLjOLZtSwaFfZxzwzAMwwDkNE1DIoDkoyhSFCXLMpR913UppVJ3ymIJg1B3EEMQMwgY51ZV9ccff3DOR6PRzc3N6ekpPGWaJlzw9PQ0m81s2z4/Px+Px77vE0Jk0a2qKoqiNE0BcwABQVIUhTRNY1kWdAIqUF3XYRjatp0kiazG2+329PT05OQE2raqKsdxptPpeDxGVkONzGazOI6x53A4RPEHUTmOA9BWVVWWZV3XhmFA3sNcmUeS5DRNc10XV33//n1RFGBT13UPh0Nd113XnZycoGowxgaDwTfffOP7vhCCvy1gfrPZqKqapulwOATHtW1rGEbXdQS1sGkasKCiKEVRUEpBhEhsWAOcp2nadV0YhhcXF+PxeDQafWVXQgghtm2v1+ssy6Ioko3KMVMieihjEpnwEec8y7LxeIwvImuQC23b/vTTT4wxqIC6roUQjuMEQbDf7znnXdeNx+OLi4swDIuiSNOUEALhLIRI03S1Wn38+BGmCiEOhwNjDEgnlmVRSvf7PYJvGAZKaF3XqFtQRFEUQbemaUopvby8fP/+vWSjvu+rqhqPx1EUGYax3++RBXAioo2+yHhbwCTnPE1TZHhRFHEcoyNG8GV/AdA6jlOW5WKxeHl5SdMU6XZ6eopoj0ajwWAghFitVsvlcr/fQxFMJpMsy56fny8uLhzHGQwGnPOiKA6HAwQh6fseQlrTNErp4XCA10HIXdcxxkaj0bt374Cfh4eH29tbeHS9XuNiTdOMRiMITzhuuVxeXFxgT+SXbAdeXl4syzIMA2j866+/wHBw0GKx+PTpE74CUECStm2LFn4+n282G1Sf7XZbVdVut/M87/z83DTNzWaj6/qPP/6IUz5//lyWZdu26Jonk0mapqgvSEDOOQE9Ah7QHpeXl8PhUNO0/X4PoE4mk/F4vNvtXl5erq6uwjDc7Xar1Wo+n0O6q6p6fX09nU6jKCKEoHsDmKGydrsdKqiu6w8PD57n3dzcSBmnvDXRgDRGCZJ9+7c5haIoo9Hohx9+uL6+hjviOP77779xpcPhUJYlYNj3PUA6HA7rup7NZl++fNnv99i56zrZCnxtk5B7kKgYToxGI0ppkiQo1GiQl8vlZrP55ZdfPM9bLpevr695nqMm2bY9n88ppb7vY+4iL4CcxNwCpXS73QIRsitT3gYN+BM8L9sQ/AtUoijKYDCIokgIkWWZbPuzLLu/v6/rejqd2rYNL8RxjHb45uYGYcdZQJ8U1ESOGY5HV7h5EASWZZVlGcfxcrlcLBbQ0bhGHMfoL8qyBKKKooBqAKRlRwAr0bpBR0ACcc6Vo4mD+jahkfX8eDoihyvgzizLsiyrqsp1XVypLMvX19csy8IwnEwmKEnQQkEQmKb5+PiIdJM7I8XIcbhQ3sqyPBwOOLIsy+12+/r6+vr6KoS4vr5GFWWMATJwHyoiRAtuZZomPkGpF0IAHcfdtGwf5fgQ94eahKqBbVJEtG3LGNtsNrPZLEmStm3zPPd9PwzDpmnKstzv94A9ZMXJyUnXdWiNYIOcKUGkaJr275gBizE2m812ux1UHVQqstRxnNFohI7IcRzLsuSQE7xgGEYQBFmWFUXheR7c+v84R8ZxzhFbGV64AyIfZIEOSs4UpJ1wRJZldV2fn59fXV0JIWazWV3Xvu83TYOC37btfr+HCoRShpF93yMesOHfSTHwxhjb7Xaz2QxMBjuQEUjFPM81TUPjuN1uUagVRUEt7PsekJH0DslwPM9AQcJA2bZtx3HknAbBh+CX4w2ZnjAVSWTb9tnZGWPsw4cPURQdDgc8ABsYY6ZpFkWBGUHTNIir7/unp6do4TRNK4pCURQCBEL5WkeLECIxj1pSlmWWZWdnZ67rIsiIhmmaXdd99913EG1PT0+73Q6X55zLeTziIEc3KPKUUs/zUKJluVbeZsHII3hBznk0TXMc5+rqajQaoUFGgwDlAuEkG18ctFgs7u7uVFX99OkTdgNVw8Wk7/urq6vr62vXdTVN831/MpkURQGbCCFRFKHmh2GoKArEkmEYNzc3GL2Cw8/Ozubz+d3dHTpczO0gumXDI10AcSqEKIri9PQUkxjARLZY6AUZY67rZlkWBAHSWDoCEYI6+PPPP/GyAMNylKSTk5M8zz9//vzPP/+UZTkej5+enr7//nvLsoqiKIriqy749ddfwzDEiA77np+fw3Q4HhZjcBCGYRiGUPIIkZT3T09P2+32cDhIoS0pBuBHuqEQoPtCt7Pb7aDw8YykdyHEZrPZbre3t7dBEKARRMelHb1NuLu7S9O0LMskSVarFTiLcx7H8Xa7Xa1WIKzRaNR13cPDAxokEBPUMJGTItiNHqbrOpQNlDRJVIfDQaIXnVVRFEmSlGUJTzPG4DiUUgxU0IYeT5CEEPf391VVmaa5XC4hSSFX4zj+8uULZOx8Pody+/Dhw3a7Vd/mooAJiDNJEuTdYrHYbDZhGOJoTE0whpKwEkI8PDxAa2JW6DgOQW/rui6KZF3XiBIir7wN1CHXOOeLxQIpQymFRkySJEkSvK6QCVbX9eFwQK8O3kb05ERgPp8nSQIO7vseAwhc+OnpabFYYIJc1/Xj46MMOzpfOUrDnG69Xruuiz4dwxi0RrZto9AwxuBuxEx25ei7iG3bXdclSQKhjhmW67q4D4ollhT5TdOs1+uyLCU5g2DleAPOwhgCEBBHL4uAKUop5uiUUogIqOMoiuI4hveBC7Tqk8kEOJJjWCTp2dnZarWS776wMEFEuskuVgiBxgl9fdu2XysCkIlwwfTjuQ3goB7NxSEH8P0syzBcQS+gHL3hkFVQHL1rOda/cvSI+6OnwOaKoqCeYQccNBgMjt+yKm+vz6B/4jgG/uM4Ro/rui4+AWCPTYJHICjLsiQyCeVoGbkqBwRSxktZgk3R7YG0JECOhb2UG/LlhyR/eWdABhILC90EuiB4qq7rNE3BvgivbOogycbjMTo6x3GapvE8D/tAPqAGg90cxwHMpQxt2/Z/mObdP3z9KOoAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=86x31 at 0x7F363E6C3470>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = Pordliest\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=118x31 at 0x7F363E772278>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "predicted text = SAWING\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gt7LDgxoaw3Y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "## results can be improved with finetuning and including more data :)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}