augment-handwritten-digit-images.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "31ff0bd2",
   "metadata": {
    "papermill": {
     "duration": 0.016372,
     "end_time": "2022-05-11T18:46:48.256073",
     "exception": false,
     "start_time": "2022-05-11T18:46:48.239701",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### **This notebook is a simple attempt on how to augment an image dataset. I will be using the [Bengali handwritten digit dataset](https://www.kaggle.com/datasets/wchowdhu/bengali-digits) for augmentation.**\n",
    "\n",
    "#### Let's get started!"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8b5efad6",
   "metadata": {
    "papermill": {
     "duration": 0.014807,
     "end_time": "2022-05-11T18:46:48.286302",
     "exception": false,
     "start_time": "2022-05-11T18:46:48.271495",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Introduction"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a5d9bd9e",
   "metadata": {
    "papermill": {
     "duration": 0.01504,
     "end_time": "2022-05-11T18:46:48.316415",
     "exception": false,
     "start_time": "2022-05-11T18:46:48.301375",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The amount of labeled data available to train a machine learning model might impact the model’s performance. This is especially true in case of deep learning-based models that generally benefit from larger amounts of annotated training examples to be able to distinguish between the different output classes. However it can be an expensive and time-consuming process to manually annotate additional data. Data augmentation can be used to generate synthetic data using the original data we have.\n",
    "\n",
    "In this tutorial, I show how to use basic transformations on input digit images to increase a training dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8a404687",
   "metadata": {
    "id": "OrF9lsDDIUqx",
    "papermill": {
     "duration": 0.014725,
     "end_time": "2022-05-11T18:46:48.346182",
     "exception": false,
     "start_time": "2022-05-11T18:46:48.331457",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Import the Necessary Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "4c38192e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-05-11T18:46:48.379475Z",
     "iopub.status.busy": "2022-05-11T18:46:48.378387Z",
     "iopub.status.idle": "2022-05-11T18:46:50.206658Z",
     "shell.execute_reply": "2022-05-11T18:46:50.205645Z"
    },
    "id": "g0w5dOaPLEz2",
    "papermill": {
     "duration": 1.8478,
     "end_time": "2022-05-11T18:46:50.209235",
     "exception": false,
     "start_time": "2022-05-11T18:46:48.361435",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import glob\n",
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import sys\n",
    "import random\n",
    "from tqdm.notebook import trange\n",
    "import skimage.io as io\n",
    "from skimage.transform import rotate, rescale, resize\n",
    "from skimage.util import random_noise\n",
    "from skimage import exposure\n",
    "from skimage.filters import gaussian\n",
    "from skimage.io import imread, imsave\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9c0a5cc0",
   "metadata": {
    "id": "Em9oebD_JBLr",
    "papermill": {
     "duration": 0.014905,
     "end_time": "2022-05-11T18:46:50.239482",
     "exception": false,
     "start_time": "2022-05-11T18:46:50.224577",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Read the Data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cfb174b1",
   "metadata": {
    "id": "KXx6qjRQI5Gs",
    "papermill": {
     "duration": 0.014873,
     "end_time": "2022-05-11T18:46:50.269606",
     "exception": false,
     "start_time": "2022-05-11T18:46:50.254733",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Read the data files from the source directory and assign corresponding target labels to the images. The image files and corresponding labels are then loaded into a pandas dataframe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "7027c088",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-05-11T18:46:50.302083Z",
     "iopub.status.busy": "2022-05-11T18:46:50.301192Z",
     "iopub.status.idle": "2022-05-11T18:46:52.795772Z",
     "shell.execute_reply": "2022-05-11T18:46:52.794656Z"
    },
    "id": "ctKpruhKYobs",
    "outputId": "7c244473-e3c7-4bcf-99b8-813080b4f4af",
    "papermill": {
     "duration": 2.51373,
     "end_time": "2022-05-11T18:46:52.798571",
     "exception": false,
     "start_time": "2022-05-11T18:46:50.284841",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>file</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>../input/bengali-digits/bengali_digits/0/1_DHA...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>../input/bengali-digits/bengali_digits/0/1_MAD...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>../input/bengali-digits/bengali_digits/0/1_DHA...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>../input/bengali-digits/bengali_digits/0/1_NOA...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>../input/bengali-digits/bengali_digits/0/1_B.B...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                file  label\n",
       "0  ../input/bengali-digits/bengali_digits/0/1_DHA...      0\n",
       "1  ../input/bengali-digits/bengali_digits/0/1_MAD...      0\n",
       "2  ../input/bengali-digits/bengali_digits/0/1_DHA...      0\n",
       "3  ../input/bengali-digits/bengali_digits/0/1_NOA...      0\n",
       "4  ../input/bengali-digits/bengali_digits/0/1_B.B...      0"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def read_data(data_dir='../input/bengali-digits/bengali_digits'):\n",
    "    ''' Read data from source directory '''\n",
    "    data = []\n",
    "    labels = []\n",
    "    for data_type in ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']:\n",
    "        digit = int(data_type[-1])\n",
    "        path = os.path.join(data_dir, data_type, '*.jpg')\n",
    "        files = glob.glob(path)\n",
    "        for f in files:\n",
    "            data.append(f)\n",
    "            labels.append(digit)\n",
    "    assert len(data) == len(labels), \"{}/{} Data size does not match labels size\".format(len(data), len(labels))\n",
    "    return data, labels\n",
    "\n",
    "\n",
    "source_dir = '../input/bengali-digits/bengali_digits'\n",
    "data, labels = read_data(source_dir)\n",
    "assert len(data) == len(labels), \"Number of image files do not match number of labels!\"\n",
    "\n",
    "df = pd.DataFrame(list(zip(data, labels)), columns =['file', 'label'])\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "277557db",
   "metadata": {
    "id": "SNMYan0jJNo1",
    "papermill": {
     "duration": 0.015497,
     "end_time": "2022-05-11T18:46:52.830300",
     "exception": false,
     "start_time": "2022-05-11T18:46:52.814803",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Load and Split the Data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3f315a8d",
   "metadata": {
    "papermill": {
     "duration": 0.015315,
     "end_time": "2022-05-11T18:46:52.861185",
     "exception": false,
     "start_time": "2022-05-11T18:46:52.845870",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The dataset is split into separate training, validation, and test subsets to train and evaluate the performance of the machine learning model. Stratification is applied to preserve the original distribution of the target classes."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "88543eee",
   "metadata": {
    "papermill": {
     "duration": 0.015266,
     "end_time": "2022-05-11T18:46:52.892044",
     "exception": false,
     "start_time": "2022-05-11T18:46:52.876778",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The following preprocessing steps are applied to the input images:\n",
    "\n",
    "- Resize all the input images to a fixed shape\n",
    "- Normalize the image tensors to have pixel values scaled down between -1 and 1 from 0 to 255 which helps with faster convergence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "2fc13dd5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-05-11T18:46:52.925589Z",
     "iopub.status.busy": "2022-05-11T18:46:52.924990Z",
     "iopub.status.idle": "2022-05-11T18:48:40.504185Z",
     "shell.execute_reply": "2022-05-11T18:48:40.502881Z"
    },
    "id": "yOfZnX2MsBYd",
    "outputId": "31115d5c-fbe5-4b29-dcb2-5e7ac848b597",
    "papermill": {
     "duration": 107.598504,
     "end_time": "2022-05-11T18:48:40.506288",
     "exception": false,
     "start_time": "2022-05-11T18:46:52.907784",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 15620/15620 [01:47<00:00, 145.72it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "original train data shape (15620, 28, 28, 3) (15620,)\n",
      "Total Training Examples in Original Data: 9996\n",
      "Total Validation Examples in Original Data: 2500\n",
      "Total Test Examples in Original Data: 3124\n"
     ]
    }
   ],
   "source": [
    "def load_data(df):\n",
    "  ''' Load and process data '''\n",
    "  all_images = []\n",
    "  for img_file in tqdm(df['file'].tolist()):\n",
    "    img = imread(img_file)\n",
    "    img = img/255.0 # Normalize\n",
    "    img = resize(img, (28, 28), anti_aliasing=True)\n",
    "    all_images.append(img)\n",
    "\n",
    "  x_np = np.stack(all_images, axis=0)\n",
    "  y_np = df['label'].values\n",
    "  print('original train data shape', x_np.shape, y_np.shape)\n",
    "  return x_np, y_np\n",
    "\n",
    "all_images, all_labels = load_data(df)\n",
    "\n",
    "all_train_features, test_features, all_train_labels, test_labels = train_test_split(all_images, all_labels, test_size=0.2, random_state=123, stratify=all_labels)\n",
    "train_features, val_features, train_labels, val_labels = train_test_split(all_train_features, all_train_labels, test_size=0.2, random_state=123, stratify=all_train_labels)\n",
    "\n",
    "print(\"Total Training Examples in Original Data: {}\".format(len(train_features)))\n",
    "print(\"Total Validation Examples in Original Data: {}\".format(len(val_features)))\n",
    "print(\"Total Test Examples in Original Data: {}\".format(len(test_features)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1b9d9752",
   "metadata": {
    "id": "xfg0x8vBJRS9",
    "papermill": {
     "duration": 0.380193,
     "end_time": "2022-05-11T18:48:41.264944",
     "exception": false,
     "start_time": "2022-05-11T18:48:40.884751",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Apply Data Augmentation "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2714ca94",
   "metadata": {
    "papermill": {
     "duration": 0.385739,
     "end_time": "2022-05-11T18:48:42.029606",
     "exception": false,
     "start_time": "2022-05-11T18:48:41.643867",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Data augmentation will only be applied on the training set to increase the number of training samples. Since data consists of digit images, I apply transformations to an input image that will preserve the original digit number. Hence 4 different transformations are applied to each input image:\n",
    "\n",
    "- Blur the image\n",
    "- Rotate the image to 25 degrees retaining the digit shape\n",
    "- Apply noise to the image\n",
    "- Apply brightness to the image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f89c7d52",
   "metadata": {
    "papermill": {
     "duration": 0.377059,
     "end_time": "2022-05-11T18:48:42.783008",
     "exception": false,
     "start_time": "2022-05-11T18:48:42.405949",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Python's image processing library, `skimage`, is used to generate the new transformed images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "aebc42e1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-05-11T18:48:43.541840Z",
     "iopub.status.busy": "2022-05-11T18:48:43.541547Z",
     "iopub.status.idle": "2022-05-11T18:48:54.458111Z",
     "shell.execute_reply": "2022-05-11T18:48:54.457235Z"
    },
    "id": "1EGjryj2pOpw",
    "outputId": "950363da-de63-4845-91b7-67b655a40d7d",
    "papermill": {
     "duration": 11.298012,
     "end_time": "2022-05-11T18:48:54.461946",
     "exception": false,
     "start_time": "2022-05-11T18:48:43.163934",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "04bdad40f2c64edb9591669aed8183da",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/9996 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:8: FutureWarning: `multichannel` is a deprecated argument name for `gaussian`. It will be removed in version 1.0. Please use `channel_axis` instead.\n",
      "  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Training Examples after Augmentation: 49980\n"
     ]
    }
   ],
   "source": [
    "def augment_data(train_features, train_labels):\n",
    "  ''' Augment data by applying different transformations '''\n",
    "  augmented_train_features, augmented_train_labels = [], []\n",
    "  train_size = len(train_features)\n",
    "  for i in trange(train_size):\n",
    "    image = train_features[i]\n",
    "    augmented_train_features.append(image) #original image\n",
    "    augmented_train_features.append(gaussian(image, sigma=1, multichannel=True)) #blurred image\n",
    "    augmented_train_features.append(rotate(image, angle=25, mode='wrap')) #rotated image\n",
    "    augmented_train_features.append(random_noise(image, var=0.2**2)) #noisy image\n",
    "    augmented_train_features.append(exposure.adjust_gamma(image, 2)) #dark image\n",
    "    for j in range(5):\n",
    "      augmented_train_labels.append(train_labels[i])\n",
    "  return np.stack(augmented_train_features, axis=0), augmented_train_labels\n",
    "\n",
    "\n",
    "\n",
    "augmented_train_features, augmented_train_labels = augment_data(train_features, train_labels)\n",
    "print(\"Total Training Examples after Augmentation: {}\".format(len(augmented_train_features)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "921cd4a3",
   "metadata": {
    "id": "WqJI0opOJopR",
    "papermill": {
     "duration": 0.378823,
     "end_time": "2022-05-11T18:48:55.222796",
     "exception": false,
     "start_time": "2022-05-11T18:48:54.843973",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "The training set increased from 9,996 examples to 49,980 images. The augmented training dataset has 49,980 images containing different transformations of the original training images."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "536a615a",
   "metadata": {
    "papermill": {
     "duration": 0.379821,
     "end_time": "2022-05-11T18:48:55.983492",
     "exception": false,
     "start_time": "2022-05-11T18:48:55.603671",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Visualize the Applied Transformations to an Input Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "06b428db",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-05-11T18:48:56.745114Z",
     "iopub.status.busy": "2022-05-11T18:48:56.744826Z",
     "iopub.status.idle": "2022-05-11T18:48:57.058491Z",
     "shell.execute_reply": "2022-05-11T18:48:57.057585Z"
    },
    "id": "03ct1rM3RddY",
    "outputId": "bc0184ef-19f8-4e35-b325-b5a3d427d101",
    "papermill": {
     "duration": 0.698662,
     "end_time": "2022-05-11T18:48:57.060630",
     "exception": false,
     "start_time": "2022-05-11T18:48:56.361968",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x1800 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(nrows=1, ncols=5, figsize=(25, 25))\n",
    "for i in range(5):\n",
    "    ax[i].imshow(augmented_train_features[i])\n",
    "    ax[i].axis('off')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aa9a2bb0",
   "metadata": {
    "papermill": {
     "duration": 0.380139,
     "end_time": "2022-05-11T18:48:57.821766",
     "exception": false,
     "start_time": "2022-05-11T18:48:57.441627",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "That's it! This is a small tutorial for beginners to get started on data augmentation for digit images.\n",
    "\n",
    "For a tutorial on implementing a classification model for Bengali handwritten digit recognition, you can check this [notebook](https://www.kaggle.com/code/wchowdhu/tutorial-on-bengali-digit-classification).\n",
    "\n",
    "\n",
    "\n",
    "<span style=\"color:blue\">If you found this notebook helpful, please leave your feedback or any suggestions, and kindly upvote. Thanks!</span>"
   ]
  }
 ],
 "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": "3.7.12"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 140.81884,
   "end_time": "2022-05-11T18:48:59.030198",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2022-05-11T18:46:38.211358",
   "version": "2.3.4"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {
     "04bdad40f2c64edb9591669aed8183da": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_61e0d8b1812f44269b4800d1e9844222",
        "IPY_MODEL_8488e2e0defc4877beb07fa63df2e686",
        "IPY_MODEL_c7c1a1be4d7a47fbb4b025460bf09c69"
       ],
       "layout": "IPY_MODEL_60c23741cee94154acca5c13d73c0403"
      }
     },
     "0b64889b72c54f4eb390f3ce3a2c4b13": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "1f4dd331ec1949fe8e99788d9b1a0c61": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "5cdeaa17de0e490fb0e5dce0061a669a": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "60c23741cee94154acca5c13d73c0403": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "61e0d8b1812f44269b4800d1e9844222": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_1f4dd331ec1949fe8e99788d9b1a0c61",
       "placeholder": "​",
       "style": "IPY_MODEL_ed7ba45f28354807a25c7e262405f94c",
       "value": "100%"
      }
     },
     "8488e2e0defc4877beb07fa63df2e686": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_c2397129905b4531a07016b7a7c232f4",
       "max": 9996.0,
       "min": 0.0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_0b64889b72c54f4eb390f3ce3a2c4b13",
       "value": 9996.0
      }
     },
     "a375377fb828439d9cc0bf1f8aef8cfb": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "DescriptionStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "DescriptionStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "description_width": ""
      }
     },
     "c2397129905b4531a07016b7a7c232f4": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "c7c1a1be4d7a47fbb4b025460bf09c69": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_5cdeaa17de0e490fb0e5dce0061a669a",
       "placeholder": "​",
       "style": "IPY_MODEL_a375377fb828439d9cc0bf1f8aef8cfb",
       "value": " 9996/9996 [00:10&lt;00:00, 955.81it/s]"
      }
     },
     "ed7ba45f28354807a25c7e262405f94c": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "DescriptionStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "DescriptionStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "description_width": ""
      }
     }
    },
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}