Test_HPopt-Attn-CTC.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [
        "s-Wwk3MyDZoj"
      ],
      "toc_visible": true,
      "authorship_tag": "ABX9TyNhC1ZFQncb929aqj0Rf8MA",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/bdu-birhanu/HHD-Ethiopic/blob/main/demo_Plain_CTC.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Download the Test Data  and Model\n",
        " This section demonstrates the process of downloading the test data  and model and evaluating its performance. Follow the steps below:"
      ],
      "metadata": {
        "id": "V0MqYzYe_bn8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import editdistance\n",
        "import tensorflow as tf\n",
        "from keras import backend as K\n",
        "import cv2\n"
      ],
      "metadata": {
        "id": "M8iG-5zQCCui"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Np669i8l8tC8"
      },
      "outputs": [],
      "source": [
        "from google.colab import drive\n",
        "import requests\n",
        "import zipfile\n",
        "import os\n",
        "# Mount your Google Drive\n",
        "drive.mount('/content/drive')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Download the Testset\n",
        "Use the following code to download the test data ( data randomly selectd from the training set)from the Hugging face. Please follow the same step if your are interested withe second test dataset.  And extrct them in to a path as the dataset is in zip file."
      ],
      "metadata": {
        "id": "luuDUfPWA_-p"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import requests\n",
        "import zipfile\n",
        "import numpy as np\n",
        "\n",
        "# Download the zip file from Hugging Face (when coping ht elink just replace \"blob\" by \"resolve\" so as to reference the file contents.)\n",
        "url= \"https://huggingface.co/datasets/OCR-Ethiopic/HHD-Ethiopic/resolve/main/test/test_rand/test_rand_numpy.zip\"\n",
        "response = requests.get(url)\n",
        "\n",
        "# Save the zip file\n",
        "zip_path = \"/content/test_rand_numpy.zip\"\n",
        "with open(zip_path, \"wb\") as file:\n",
        "    file.write(response.content)\n",
        "\n",
        "# Extract the zip file\n",
        "extract_path = \"/content/test\"\n",
        "with zipfile.ZipFile(zip_path, \"r\") as zip_ref:\n",
        "    zip_ref.extractall(extract_path)\n",
        "\n",
        "# Load the numpy file from the path we extracted\n",
        "import numpy as np\n",
        "x_test_rand= np.load('/content/test/test_rand_numpy/x_test_rand.npy',allow_pickle=True)\n",
        "y_test_rand= np.load('/content/test/test_rand_numpy/y_test_rand.npy',allow_pickle=True)"
      ],
      "metadata": {
        "id": "BQ6pVu69OKxo"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_test_rand.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tC7WI3wHPeaP",
        "outputId": "ff808208-00e0-4ea7-a581-d11ee3a15f95"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(6375,)"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "### Image size normalization\n",
        "Normalize the image to 48 by 368, keeping the aspect-ratio."
      ],
      "metadata": {
        "id": "s-Wwk3MyDZoj"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "im_row=48\n",
        "im_col=368\n",
        "def resize(x_image):\n",
        "\n",
        "\n",
        "    resized_x_rand = np.zeros((len(x_image), im_row, im_col), dtype=np.uint8)\n",
        "\n",
        "    # loop over the input images\n",
        "    for i, image in enumerate(x_image):\n",
        "        # resize the image to 48 by 368 using padding\n",
        "        current_height, current_width = image.shape[:2]\n",
        "        aspect_ratio_current = current_width / current_height\n",
        "        aspect_ratio_target = im_col / im_row\n",
        "        if aspect_ratio_current != aspect_ratio_target:\n",
        "            if aspect_ratio_current > aspect_ratio_target:\n",
        "                new_height = int(current_width / aspect_ratio_target)\n",
        "                top_padding = (new_height - current_height) // 2\n",
        "                bottom_padding = new_height - current_height - top_padding\n",
        "                padded_image = cv2.copyMakeBorder(image, top_padding, bottom_padding, 0, 0, cv2.BORDER_CONSTANT, value=255)\n",
        "            else:\n",
        "                new_width = int(current_height * aspect_ratio_target)\n",
        "                left_padding = (new_width - current_width) // 2\n",
        "                right_padding = new_width - current_width - left_padding\n",
        "                padded_image = cv2.copyMakeBorder(image, 0, 0, 0, right_padding+left_padding, cv2.BORDER_CONSTANT, value=255)\n",
        "        else:\n",
        "            padded_image = image\n",
        "        resized_image = cv2.resize(padded_image, (im_col, im_row))\n",
        "\n",
        "        resized_x_rand[i] = resized_image\n",
        "\n",
        "    return resized_x_rand"
      ],
      "metadata": {
        "id": "xgIDxNgUC_2R"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_test_rand_resized= resize(x_test_rand)"
      ],
      "metadata": {
        "id": "_cw-4EUwDjip"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Visualizie a sample resized image"
      ],
      "metadata": {
        "id": "krVunqBnJZGZ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt \n",
        "plt.imshow(x_test_rand[1400], cmap='gray')\n",
        "plt.xticks([])\n",
        "plt.yticks([])\n",
        "plt.gca().spines['left'].set_visible(True)  # Show left border line\n",
        "plt.gca().spines['right'].set_visible(True)  # Hide right border line\n",
        "plt.gca().spines['top'].set_visible(True)  # Hide top border line\n",
        "plt.gca().spines['bottom'].set_visible(True)  # Show bottom border line\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "EX0Qt7u-ofFr",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 79
        },
        "outputId": "43d74167-c3a6-42cd-a6f4-25d0ad11b597"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        " #dictionary used to decode the numeric predictions to text\n",
        "dict_eth={1: 'ሀ', 2: 'ሁ', 3: 'ሂ', 4: 'ሃ', 5: 'ሄ', 6: 'ህ', 7: 'ሆ', 8: 'ለ', 9: 'ሉ', 10: 'ሊ', 11: 'ላ', 12: 'ሌ', 13: 'ል', 14: 'ሎ', 15: 'ሏ', 16: 'ሐ', 17: 'ሑ', 18: 'ሒ', 19: 'ሓ', 20: 'ሔ', 21: 'ሕ', 22: 'ሖ', 23: 'መ', 24: 'ሙ', 25: 'ሚ', 26: 'ማ', 27: 'ሜ', 28: 'ም', 29: 'ሞ', 30: 'ሟ', 31: 'ሠ', 32: 'ሡ', 33: 'ሢ', 34: 'ሣ', 35: 'ሤ', 36: 'ሥ', 37: 'ሦ', 38: 'ረ', 39: 'ሩ', 40: 'ሪ', 41: 'ራ', 42: 'ሬ', 43: 'ር', 44: 'ሮ', 45: 'ሯ', 46: 'ሰ', 47: 'ሱ', 48: 'ሲ', 49: 'ሳ', 50: 'ሴ', 51: 'ስ', 52: 'ሶ', 53: 'ሷ', 54: 'ሸ', 55: 'ሹ', 56: 'ሺ', 57: 'ሻ', 58: 'ሼ', 59: 'ሽ', 60: 'ሾ', 61: 'ሿ', 62: 'ቀ', 63: 'ቁ', 64: 'ቂ', 65: 'ቃ', 66: 'ቄ', 67: 'ቅ', 68: 'ቆ', 69: 'ቈ', 70: 'ቊ', 71: 'ቋ', 72: 'ቍ', 73: 'በ', 74: 'ቡ', 75: 'ቢ', 76: 'ባ', 77: 'ቤ', 78: 'ብ', 79: 'ቦ', 80: 'ቧ', 81: 'ቨ', 82: 'ቩ', 83: 'ቪ', 84: 'ቫ', 85: 'ቬ', 86: 'ቭ', 87: 'ቮ', 88: 'ተ', 89: 'ቱ', 90: 'ቲ', 91: 'ታ', 92: 'ቴ', 93: 'ት', 94: 'ቶ', 95: 'ቷ', 96: 'ቸ', 97: 'ቹ', 98: 'ቺ', 99: 'ቻ', 100: 'ቼ', 101: 'ች', 102: 'ቾ', 103: 'ቿ', 104: 'ኀ', 105: 'ኁ', 106: 'ኂ', 107: 'ኃ', 108: 'ኄ', 109: 'ኅ', 110: 'ኆ', 111: 'ኈ', 112: 'ኊ', 113: 'ኋ', 114: 'ኍ', 115: 'ነ', 116: 'ኑ', 117: 'ኒ', 118: 'ና', 119: 'ኔ', 120: 'ን', 121: 'ኖ', 122: 'ኗ', 123: 'ኘ', 124: 'ኙ', 125: 'ኚ', 126: 'ኛ', 127: 'ኜ', 128: 'ኝ', 129: 'ኞ', 130: 'ኟ', 131: 'አ', 132: 'ኡ', 133: 'ኢ', 134: 'ኣ', 135: 'ኤ', 136: 'እ', 137: 'ኦ', 138: 'ኧ', 139: 'ከ', 140: 'ኩ', 141: 'ኪ', 142: 'ካ', 143: 'ኬ', 144: 'ክ', 145: 'ኮ', 146: 'ኰ', 147: 'ኲ', 148: 'ኳ', 149: 'ኴ', 150: 'ኵ', 151: 'ኸ', 152: 'ኹ', 153: 'ኺ', 154: 'ኻ', 155: 'ኼ', 156: 'ኽ', 157: 'ኾ', 158: 'ወ', 159: 'ዉ', 160: 'ዊ', 161: 'ዋ', 162: 'ዌ', 163: 'ው', 164: 'ዎ', 165: 'ዏ', 166: 'ዐ', 167: 'ዑ', 168: 'ዒ', 169: 'ዓ', 170: 'ዔ', 171: 'ዕ', 172: 'ዖ', 173: 'ዘ', 174: 'ዙ', 175: 'ዚ', 176: 'ዛ', 177: 'ዜ', 178: 'ዝ', 179: 'ዞ', 180: 'ዟ', 181: 'ዠ', 182: 'ዡ', 183: 'ዢ', 184: 'ዣ', 185: 'ዤ', 186: 'ዥ', 187: 'ዦ', 188: 'የ', 189: 'ዩ', 190: 'ዪ', 191: 'ያ', 192: 'ዬ', 193: 'ይ', 194: 'ዮ', 195: 'ደ', 196: 'ዱ', 197: 'ዲ', 198: 'ዳ', 199: 'ዴ', 200: 'ድ', 201: 'ዶ', 202: 'ዷ', 203: 'ዿ', 204: 'ጀ', 205: 'ጁ', 206: 'ጂ', 207: 'ጃ', 208: 'ጄ', 209: 'ጅ', 210: 'ጆ', 211: 'ጇ', 212: 'ገ', 213: 'ጉ', 214: 'ጊ', 215: 'ጋ', 216: 'ጌ', 217: 'ግ', 218: 'ጎ', 219: 'ጐ', 220: 'ጒ', 221: 'ጓ', 222: 'ጔ', 223: 'ጕ', 224: 'ጠ', 225: 'ጡ', 226: 'ጢ', 227: 'ጣ', 228: 'ጤ', 229: 'ጥ', 230: 'ጦ', 231: 'ጧ', 232: 'ጨ', 233: 'ጩ', 234: 'ጪ', 235: 'ጫ', 236: 'ጬ', 237: 'ጭ', 238: 'ጮ', 239: 'ጯ', 240: 'ጰ', 241: 'ጱ', 242: 'ጲ', 243: 'ጳ', 244: 'ጴ', 245: 'ጵ', 246: 'ጶ', 247: 'ጸ', 248: 'ጹ', 249: 'ጺ', 250: 'ጻ', 251: 'ጼ', 252: 'ጽ', 253: 'ጾ', 254: 'ጿ', 255: 'ፀ', 256: 'ፁ', 257: 'ፂ', 258: 'ፃ', 259: 'ፄ', 260: 'ፅ', 261: 'ፆ', 262: 'ፈ', 263: 'ፉ', 264: 'ፊ', 265: 'ፋ', 266: 'ፌ', 267: 'ፍ', 268: 'ፎ', 269: 'ፏ', 270: 'ፐ', 271: 'ፑ', 272: 'ፒ', 273: 'ፓ', 274: 'ፔ', 275: 'ፕ', 276: 'ፖ', 277: 'ፚ', 278: '፠', 279: '፡', 280: '።', 281: '፣', 282: '፤', 283: '፥', 284: '፦', 285: '፨', 286: '፩', 287: '፪', 288: '፫', 289: '፬', 290: '፭', 291: '፮', 292: '፯', 293: '፰', 294: '፱', 295: '፲', 296: '፳', 297: '፴', 298: '፵', 299: '፶', 300: '፷', 301: '፸', 302: '፹', 303: '፺', 304: '፻', 305: '፼', 306: ' '}"
      ],
      "metadata": {
        "id": "soveRiql-1Nq"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "k=[]\n",
        "for i,j in dict_eth.items():\n",
        "  k.append(j)\n",
        "print(k)"
      ],
      "metadata": {
        "id": "xZ3LKjEwh-B7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Load the pretrained Model:\n",
        " After downloading the test data and preporfces it, load the one of the pretrained Ethiopic OCR model from github repository as follows."
      ],
      "metadata": {
        "id": "XXuF_2boCop1"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "%pip install wget\n",
        "import wget\n",
        "import tensorflow as tf\n",
        "\n",
        "m_url = \"https://github.com/bdu-birhanu/HHD-Ethiopic/raw/main/hhd-ethiopic/model_belened_CTC_deephyper_0.hdf5\"\n",
        "m_path = \"model_belened_CTC_deephyper_0.hdf5\"\n",
        "# Download the model file\n",
        "wget.download(m_url, m_path)\n",
        "\n",
        "model = tf.keras.models.load_model(m_path, custom_objects={'Functional': tf.keras.models.Model})\n",
        "\n",
        "model.summary()\n",
        "# model= tf.keras.models.load_model(\n",
        "#     '/content/drive/MyDrive/Code_2023/model/model_belened_CTC_deephyper_0.hdf5',\n",
        "#     custom_objects={'Functional':tf.keras.models.Model})\n",
        "\n",
        "# model.summary()"
      ],
      "metadata": {
        "id": "j0c7BufccYKG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Visualize the Results\n",
        " Use the loaded model and test data to generate predictions."
      ],
      "metadata": {
        "id": "RCEzmFsZETCQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y_pred=model.predict(x_test_rand_resized)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3eHtafkmCqSe",
        "outputId": "e5ca0a89-46b4-4758-97cd-5afbeeb17ba8"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "200/200 [==============================] - 19s 17ms/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### CTC Decoder\n",
        "\n",
        "employing the CTC decoder to generate the actual text  using the probablities collcted from the model prediction"
      ],
      "metadata": {
        "id": "rVKQhw8sEw9T"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y_decode = K.get_value(K.ctc_decode(y_pred[:, :, :], input_length=np.ones(y_pred.shape[0]) * y_pred.shape[1])[0][0])\n"
      ],
      "metadata": {
        "id": "TGK-8WqvFfKl"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Compute Character Error Rate(CER)\n",
        " Calculate the CER to evaluate the accuracy of the model's predictions using editdistance.\n"
      ],
      "metadata": {
        "id": "lL1Ul7mbFfTG"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "true = []  # to stor value of character by removing zero which was padded previously and also this is the value of newline in the test label\n",
        "for i in range(len(y_test_rand)):\n",
        "    x = [j for j in y_test_rand[i] if j != 0]\n",
        "    true.append(x)\n",
        "\n",
        "\n",
        "pred=[]\n",
        "# to stor the predicted charcter except zerro and -1 which are padded value nad blank space predicted during testing\n",
        "for i in range(len(y_decode)):\n",
        "    x=[j for j in y_decode[i] if j not in(0,-1)]\n",
        "    pred.append(x)\n",
        "\n",
        "cer=0\n",
        "for(i,j) in zip(true,pred):\n",
        "    x=editdistance.eval(i,j)\n",
        "    cer=cer+x\n",
        "err=cer\n",
        "x=0\n",
        "for i in range(len(true)):\n",
        "    x=x+len(true[i])\n",
        "totalchar=x\n",
        "cerp=(float(err)/totalchar)*100\n",
        "print(\"The CER of random test set:\", cerp)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HoLQWoL7IDY2",
        "outputId": "94282ebd-6721-41b4-a171-db78772352e7"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The CER of random test set: 16.396747605257296\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "###  Visualize the Image Ground-Truth and Predictions\n",
        "\n",
        "To gain insights into the model's performance, it's essential to visualize the image, the ground-truth and corresponding predictions.  See th below resUlts which we provide following these steps.\n",
        "\n",
        "1. **Load the Image and Ground-Truth Data**: Loading the image data and the corresponding ground-truth labels as sescribed above. \n",
        "2. **Generate Predictions**: Use the trained model to generate predictions for the loaded image data. \n",
        "\n",
        "3. **Visualize the Image, Ground-Truth and Predictions**:  ground-truth on the left prediction on the right and the images are at the bottom respectivly."
      ],
      "metadata": {
        "id": "w8cYFfJoZxj2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "k=1005\n",
        "b=1010\n",
        "\n",
        "\n",
        "char_list_true= [[dict_eth[i] for i in text] for text in true]\n",
        "\n",
        "text_list_true = [''.join(row) for row in char_list_true]\n",
        "\n",
        "char_list_pred = [[dict_eth[i] for i in text] for text in pred]\n",
        "\n",
        "text_list_pred = [''.join(row) for row in char_list_pred]\n",
        "\n",
        "print(f'{\"Ground-truth Texts\":<30} Prediction Texts')\n",
        "for i, j in zip(text_list_true[k:b], text_list_pred[k:b]):\n",
        "\n",
        "    print(f'{(\"\".join(i)):<30} {(\"\".join(j))}')\n",
        "# Print the text\n",
        "# print('\\n'.join(text_list[:5]))\n",
        "\n",
        "for i in range(k,b):\n",
        "    fig = plt.figure(figsize=(5, 2))\n",
        "    plt.imshow(x_test_rand[i], cmap='gray')\n",
        "    plt.xticks([])\n",
        "    plt.yticks([])\n",
        "    plt.gca().spines['left'].set_visible(True)  # Show left border line\n",
        "    plt.gca().spines['right'].set_visible(True)  # Hide right border line\n",
        "    plt.gca().spines['top'].set_visible(True)  # Hide top border line\n",
        "    plt.gca().spines['bottom'].set_visible(True)  # Show bottom border line\n",
        "    #plt.axis('off')\n",
        "    plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 461
        },
        "id": "x686ko_1NFJw",
        "outputId": "73c65ffe-fa24-419a-c054-7cf65a887216"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Ground-truth Texts             Prediction Texts\n",
            "ዕል፡ቀዳማ፡ዘውእቱ፡ሕፁፃ                ዕል፡ቀዳማ፡ዘውእቱ፡ሕፁፃ\n",
            "ሰከ፡ኅበ፡ይብል፡ሐዲከ፡ዐሉይ፡             ስከ፡ኅበ፡ይብል፡ሐደስ፡ወሉይ፡\n",
            "ር፡ከመ፡ይምሐረኒ                     ር፡ከመ፡ይምሐረኒ\n",
            "ናሁ፡አነሥእ፡እዲየ፡መን                 ናሁ፡አነሥእ፡ኔዲየ፡መን\n",
            "አሚረ፡እሙንቱ፡እለ፡የ                  አሚረ፡እሙንቱ፡እለ፡\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x200 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x200 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x200 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x200 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x200 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Sample example to comput the editdistance between the ground-truth and prediction texts and then compute the CER"
      ],
      "metadata": {
        "id": "btivRC66ZeuA"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import editdistance\n",
        "x=editdistance.eval('ዓገሠ።በዝሕማም፡መሥጋ፡ ','ዓገሠ።በዝሕማም፡በሥጋ፡')\n",
        "\n",
        "print(\"gt vs model \"+ str(x))\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dy8M7lIymDEf",
        "outputId": "075f0bb7-34cf-4da2-8c87-cfff46f385f0"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "gt vs model 2\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "gt='ዓገሠ።በዝሕማም፡መሥጋ፡'\n",
        "CER=float((x))/len(gt)*100\n",
        "print(CER)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ie4oQM4gkXZ-",
        "outputId": "b970783b-b609-4348-af40-129873e32d85"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "14.285714285714285\n"
          ]
        }
      ]
    }
  ]
}