nbs/04_relabel_loop.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "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>filepath</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/imagenette/2/00710.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/imagenette/2/00063.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/imagenette/2/00506.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>data/imagenette/2/00575.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>data/imagenette/2/00136.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>270</th>\n",
       "      <td>data/imagenette/3/02578.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>271</th>\n",
       "      <td>data/imagenette/6/07713.jpg</td>\n",
       "      <td>garbage truck</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>272</th>\n",
       "      <td>data/imagenette/6/07672.jpg</td>\n",
       "      <td>garbage truck</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>273</th>\n",
       "      <td>data/imagenette/8/07255.jpg</td>\n",
       "      <td>golf ball</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>274</th>\n",
       "      <td>data/imagenette/8/07004.jpg</td>\n",
       "      <td>golf ball</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>275 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                        filepath            label\n",
       "0    data/imagenette/2/00710.jpg  cassette player\n",
       "1    data/imagenette/2/00063.jpg  cassette player\n",
       "2    data/imagenette/2/00506.jpg  cassette player\n",
       "3    data/imagenette/2/00575.jpg  cassette player\n",
       "4    data/imagenette/2/00136.jpg  cassette player\n",
       "..                           ...              ...\n",
       "270  data/imagenette/3/02578.jpg        chain saw\n",
       "271  data/imagenette/6/07713.jpg    garbage truck\n",
       "272  data/imagenette/6/07672.jpg    garbage truck\n",
       "273  data/imagenette/8/07255.jpg        golf ball\n",
       "274  data/imagenette/8/07004.jpg        golf ball\n",
       "\n",
       "[275 rows x 2 columns]"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "# df = pd.read_parquet(\"initial_samples.parquet\")\n",
    "df = pd.read_parquet(\"active_labeled.parquet\")\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:29:35.315\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mload_model\u001b[0m:\u001b[36m30\u001b[0m - \u001b[1mLoading model resnet18\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "from active_vision import ActiveLearner\n",
    "from fastai.vision.models.all import resnet18\n",
    "\n",
    "al = ActiveLearner(resnet18)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:29:36.123\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mload_dataset\u001b[0m:\u001b[36m45\u001b[0m - \u001b[1mLoading dataset from filepath and label\u001b[0m\n",
      "\u001b[32m2025-01-13 15:29:36.124\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mload_dataset\u001b[0m:\u001b[36m48\u001b[0m - \u001b[1mCreating dataloaders\u001b[0m\n",
      "\u001b[32m2025-01-13 15:29:36.489\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mload_dataset\u001b[0m:\u001b[36m59\u001b[0m - \u001b[1mCreating learner\u001b[0m\n",
      "\u001b[32m2025-01-13 15:29:36.619\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mload_dataset\u001b[0m:\u001b[36m62\u001b[0m - \u001b[1mDone. Ready to train.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "al.load_dataset(df, filepath_col=\"filepath\", label_col=\"label\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "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>filepath</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/imagenette/2/00710.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/imagenette/2/00063.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/imagenette/2/00506.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>data/imagenette/2/00575.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>data/imagenette/2/00136.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>270</th>\n",
       "      <td>data/imagenette/3/02578.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>271</th>\n",
       "      <td>data/imagenette/6/07713.jpg</td>\n",
       "      <td>garbage truck</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>272</th>\n",
       "      <td>data/imagenette/6/07672.jpg</td>\n",
       "      <td>garbage truck</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>273</th>\n",
       "      <td>data/imagenette/8/07255.jpg</td>\n",
       "      <td>golf ball</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>274</th>\n",
       "      <td>data/imagenette/8/07004.jpg</td>\n",
       "      <td>golf ball</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>275 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                        filepath            label\n",
       "0    data/imagenette/2/00710.jpg  cassette player\n",
       "1    data/imagenette/2/00063.jpg  cassette player\n",
       "2    data/imagenette/2/00506.jpg  cassette player\n",
       "3    data/imagenette/2/00575.jpg  cassette player\n",
       "4    data/imagenette/2/00136.jpg  cassette player\n",
       "..                           ...              ...\n",
       "270  data/imagenette/3/02578.jpg        chain saw\n",
       "271  data/imagenette/6/07713.jpg    garbage truck\n",
       "272  data/imagenette/6/07672.jpg    garbage truck\n",
       "273  data/imagenette/8/07255.jpg        golf ball\n",
       "274  data/imagenette/8/07004.jpg        golf ball\n",
       "\n",
       "[275 rows x 2 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "al.train_set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:29:36.791\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mlr_find\u001b[0m:\u001b[36m65\u001b[0m - \u001b[1mFinding optimal learning rate\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:29:43.895\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mlr_find\u001b[0m:\u001b[36m67\u001b[0m - \u001b[1mOptimal learning rate: 0.001737800776027143\u001b[0m\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "al.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:29:45.468\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mtrain\u001b[0m:\u001b[36m70\u001b[0m - \u001b[1mTraining for 3 epochs with learning rate: 0.001\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>3.127271</td>\n",
       "      <td>1.446365</td>\n",
       "      <td>0.436364</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.949262</td>\n",
       "      <td>1.135409</td>\n",
       "      <td>0.654545</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1.597486</td>\n",
       "      <td>0.874778</td>\n",
       "      <td>0.727273</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1.295250</td>\n",
       "      <td>0.805245</td>\n",
       "      <td>0.745455</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "al.train(epochs=3, lr=1e-3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate the model on a *labeled* evaluation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:29:58.557\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mevaluate\u001b[0m:\u001b[36m112\u001b[0m - \u001b[1mAccuracy: 93.40%\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.9340127388535032"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_df = pd.read_parquet(\"evaluation_samples.parquet\")\n",
    "# test_df = test_df.rename(columns={\"label_name\": \"label\"})\n",
    "al.evaluate(test_df, filepath_col=\"filepath\", label_col=\"label\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get the predictions from an *unlabeled* set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['data/imagenette/2/00000.jpg',\n",
       " 'data/imagenette/2/00001.jpg',\n",
       " 'data/imagenette/2/00002.jpg',\n",
       " 'data/imagenette/2/00004.jpg',\n",
       " 'data/imagenette/2/00005.jpg',\n",
       " 'data/imagenette/2/00006.jpg',\n",
       " 'data/imagenette/2/00007.jpg',\n",
       " 'data/imagenette/2/00008.jpg',\n",
       " 'data/imagenette/2/00009.jpg',\n",
       " 'data/imagenette/2/00010.jpg',\n",
       " 'data/imagenette/2/00011.jpg',\n",
       " 'data/imagenette/2/00012.jpg',\n",
       " 'data/imagenette/2/00013.jpg',\n",
       " 'data/imagenette/2/00014.jpg',\n",
       " 'data/imagenette/2/00015.jpg',\n",
       " 'data/imagenette/2/00016.jpg',\n",
       " 'data/imagenette/2/00017.jpg',\n",
       " 'data/imagenette/2/00018.jpg',\n",
       " 'data/imagenette/2/00019.jpg',\n",
       " 'data/imagenette/2/00020.jpg',\n",
       " 'data/imagenette/2/00021.jpg',\n",
       " 'data/imagenette/2/00022.jpg',\n",
       " 'data/imagenette/2/00023.jpg',\n",
       " 'data/imagenette/2/00024.jpg',\n",
       " 'data/imagenette/2/00025.jpg',\n",
       " 'data/imagenette/2/00026.jpg',\n",
       " 'data/imagenette/2/00027.jpg',\n",
       " 'data/imagenette/2/00028.jpg',\n",
       " 'data/imagenette/2/00029.jpg',\n",
       " 'data/imagenette/2/00030.jpg',\n",
       " 'data/imagenette/2/00031.jpg',\n",
       " 'data/imagenette/2/00032.jpg',\n",
       " 'data/imagenette/2/00033.jpg',\n",
       " 'data/imagenette/2/00034.jpg',\n",
       " 'data/imagenette/2/00035.jpg',\n",
       " 'data/imagenette/2/00036.jpg',\n",
       " 'data/imagenette/2/00037.jpg',\n",
       " 'data/imagenette/2/00038.jpg',\n",
       " 'data/imagenette/2/00039.jpg',\n",
       " 'data/imagenette/2/00040.jpg',\n",
       " 'data/imagenette/2/00041.jpg',\n",
       " 'data/imagenette/2/00042.jpg',\n",
       " 'data/imagenette/2/00043.jpg',\n",
       " 'data/imagenette/2/00044.jpg',\n",
       " 'data/imagenette/2/00045.jpg',\n",
       " 'data/imagenette/2/00046.jpg',\n",
       " 'data/imagenette/2/00047.jpg',\n",
       " 'data/imagenette/2/00048.jpg',\n",
       " 'data/imagenette/2/00049.jpg',\n",
       " 'data/imagenette/2/00050.jpg',\n",
       " 'data/imagenette/2/00051.jpg',\n",
       " 'data/imagenette/2/00052.jpg',\n",
       " 'data/imagenette/2/00053.jpg',\n",
       " 'data/imagenette/2/00054.jpg',\n",
       " 'data/imagenette/2/00055.jpg',\n",
       " 'data/imagenette/2/00056.jpg',\n",
       " 'data/imagenette/2/00057.jpg',\n",
       " 'data/imagenette/2/00058.jpg',\n",
       " 'data/imagenette/2/00059.jpg',\n",
       " 'data/imagenette/2/00060.jpg',\n",
       " 'data/imagenette/2/00061.jpg',\n",
       " 'data/imagenette/2/00062.jpg',\n",
       " 'data/imagenette/2/00064.jpg',\n",
       " 'data/imagenette/2/00065.jpg',\n",
       " 'data/imagenette/2/00066.jpg',\n",
       " 'data/imagenette/2/00067.jpg',\n",
       " 'data/imagenette/2/00068.jpg',\n",
       " 'data/imagenette/2/00069.jpg',\n",
       " 'data/imagenette/2/00070.jpg',\n",
       " 'data/imagenette/2/00071.jpg',\n",
       " 'data/imagenette/2/00072.jpg',\n",
       " 'data/imagenette/2/00073.jpg',\n",
       " 'data/imagenette/2/00074.jpg',\n",
       " 'data/imagenette/2/00075.jpg',\n",
       " 'data/imagenette/2/00076.jpg',\n",
       " 'data/imagenette/2/00077.jpg',\n",
       " 'data/imagenette/2/00078.jpg',\n",
       " 'data/imagenette/2/00079.jpg',\n",
       " 'data/imagenette/2/00080.jpg',\n",
       " 'data/imagenette/2/00081.jpg',\n",
       " 'data/imagenette/2/00082.jpg',\n",
       " 'data/imagenette/2/00083.jpg',\n",
       " 'data/imagenette/2/00084.jpg',\n",
       " 'data/imagenette/2/00085.jpg',\n",
       " 'data/imagenette/2/00086.jpg',\n",
       " 'data/imagenette/2/00087.jpg',\n",
       " 'data/imagenette/2/00088.jpg',\n",
       " 'data/imagenette/2/00089.jpg',\n",
       " 'data/imagenette/2/00090.jpg',\n",
       " 'data/imagenette/2/00091.jpg',\n",
       " 'data/imagenette/2/00092.jpg',\n",
       " 'data/imagenette/2/00093.jpg',\n",
       " 'data/imagenette/2/00094.jpg',\n",
       " 'data/imagenette/2/00095.jpg',\n",
       " 'data/imagenette/2/00096.jpg',\n",
       " 'data/imagenette/2/00097.jpg',\n",
       " 'data/imagenette/2/00098.jpg',\n",
       " 'data/imagenette/2/00099.jpg',\n",
       " 'data/imagenette/2/00100.jpg',\n",
       " 'data/imagenette/2/00101.jpg',\n",
       " 'data/imagenette/2/00102.jpg',\n",
       " 'data/imagenette/2/00103.jpg',\n",
       " 'data/imagenette/2/00104.jpg',\n",
       " 'data/imagenette/2/00105.jpg',\n",
       " 'data/imagenette/2/00106.jpg',\n",
       " 'data/imagenette/2/00107.jpg',\n",
       " 'data/imagenette/2/00108.jpg',\n",
       " 'data/imagenette/2/00109.jpg',\n",
       " 'data/imagenette/2/00110.jpg',\n",
       " 'data/imagenette/2/00111.jpg',\n",
       " 'data/imagenette/2/00112.jpg',\n",
       " 'data/imagenette/2/00113.jpg',\n",
       " 'data/imagenette/2/00114.jpg',\n",
       " 'data/imagenette/2/00115.jpg',\n",
       " 'data/imagenette/2/00116.jpg',\n",
       " 'data/imagenette/2/00117.jpg',\n",
       " 'data/imagenette/2/00118.jpg',\n",
       " 'data/imagenette/2/00119.jpg',\n",
       " 'data/imagenette/2/00120.jpg',\n",
       " 'data/imagenette/2/00121.jpg',\n",
       " 'data/imagenette/2/00122.jpg',\n",
       " 'data/imagenette/2/00123.jpg',\n",
       " 'data/imagenette/2/00124.jpg',\n",
       " 'data/imagenette/2/00125.jpg',\n",
       " 'data/imagenette/2/00126.jpg',\n",
       " 'data/imagenette/2/00127.jpg',\n",
       " 'data/imagenette/2/00128.jpg',\n",
       " 'data/imagenette/2/00129.jpg',\n",
       " 'data/imagenette/2/00130.jpg',\n",
       " 'data/imagenette/2/00131.jpg',\n",
       " 'data/imagenette/2/00132.jpg',\n",
       " 'data/imagenette/2/00133.jpg',\n",
       " 'data/imagenette/2/00134.jpg',\n",
       " 'data/imagenette/2/00135.jpg',\n",
       " 'data/imagenette/2/00137.jpg',\n",
       " 'data/imagenette/2/00138.jpg',\n",
       " 'data/imagenette/2/00139.jpg',\n",
       " 'data/imagenette/2/00140.jpg',\n",
       " 'data/imagenette/2/00141.jpg',\n",
       " 'data/imagenette/2/00142.jpg',\n",
       " 'data/imagenette/2/00143.jpg',\n",
       " 'data/imagenette/2/00144.jpg',\n",
       " 'data/imagenette/2/00145.jpg',\n",
       " 'data/imagenette/2/00146.jpg',\n",
       " 'data/imagenette/2/00147.jpg',\n",
       " 'data/imagenette/2/00148.jpg',\n",
       " 'data/imagenette/2/00149.jpg',\n",
       " 'data/imagenette/2/00150.jpg',\n",
       " 'data/imagenette/2/00151.jpg',\n",
       " 'data/imagenette/2/00152.jpg',\n",
       " 'data/imagenette/2/00153.jpg',\n",
       " 'data/imagenette/2/00154.jpg',\n",
       " 'data/imagenette/2/00155.jpg',\n",
       " 'data/imagenette/2/00156.jpg',\n",
       " 'data/imagenette/2/00157.jpg',\n",
       " 'data/imagenette/2/00158.jpg',\n",
       " 'data/imagenette/2/00159.jpg',\n",
       " 'data/imagenette/2/00160.jpg',\n",
       " 'data/imagenette/2/00161.jpg',\n",
       " 'data/imagenette/2/00162.jpg',\n",
       " 'data/imagenette/2/00163.jpg',\n",
       " 'data/imagenette/2/00164.jpg',\n",
       " 'data/imagenette/2/00165.jpg',\n",
       " 'data/imagenette/2/00166.jpg',\n",
       " 'data/imagenette/2/00167.jpg',\n",
       " 'data/imagenette/2/00168.jpg',\n",
       " 'data/imagenette/2/00169.jpg',\n",
       " 'data/imagenette/2/00170.jpg',\n",
       " 'data/imagenette/2/00171.jpg',\n",
       " 'data/imagenette/2/00172.jpg',\n",
       " 'data/imagenette/2/00173.jpg',\n",
       " 'data/imagenette/2/00174.jpg',\n",
       " 'data/imagenette/2/00175.jpg',\n",
       " 'data/imagenette/2/00177.jpg',\n",
       " 'data/imagenette/2/00178.jpg',\n",
       " 'data/imagenette/2/00179.jpg',\n",
       " 'data/imagenette/2/00180.jpg',\n",
       " 'data/imagenette/2/00181.jpg',\n",
       " 'data/imagenette/2/00182.jpg',\n",
       " 'data/imagenette/2/00183.jpg',\n",
       " 'data/imagenette/2/00184.jpg',\n",
       " 'data/imagenette/2/00185.jpg',\n",
       " 'data/imagenette/2/00186.jpg',\n",
       " 'data/imagenette/2/00187.jpg',\n",
       " 'data/imagenette/2/00188.jpg',\n",
       " 'data/imagenette/2/00189.jpg',\n",
       " 'data/imagenette/2/00190.jpg',\n",
       " 'data/imagenette/2/00191.jpg',\n",
       " 'data/imagenette/2/00192.jpg',\n",
       " 'data/imagenette/2/00193.jpg',\n",
       " 'data/imagenette/2/00194.jpg',\n",
       " 'data/imagenette/2/00196.jpg',\n",
       " 'data/imagenette/2/00197.jpg',\n",
       " 'data/imagenette/2/00198.jpg',\n",
       " 'data/imagenette/2/00199.jpg',\n",
       " 'data/imagenette/2/00200.jpg',\n",
       " 'data/imagenette/2/00201.jpg',\n",
       " 'data/imagenette/2/00202.jpg',\n",
       " 'data/imagenette/2/00203.jpg',\n",
       " 'data/imagenette/2/00204.jpg',\n",
       " 'data/imagenette/2/00205.jpg',\n",
       " 'data/imagenette/2/00206.jpg',\n",
       " 'data/imagenette/2/00207.jpg',\n",
       " 'data/imagenette/2/00208.jpg',\n",
       " 'data/imagenette/2/00209.jpg',\n",
       " 'data/imagenette/2/00210.jpg',\n",
       " 'data/imagenette/2/00211.jpg',\n",
       " 'data/imagenette/2/00212.jpg',\n",
       " 'data/imagenette/2/00213.jpg',\n",
       " 'data/imagenette/2/00214.jpg',\n",
       " 'data/imagenette/2/00215.jpg',\n",
       " 'data/imagenette/2/00216.jpg',\n",
       " 'data/imagenette/2/00217.jpg',\n",
       " 'data/imagenette/2/00218.jpg',\n",
       " 'data/imagenette/2/00219.jpg',\n",
       " 'data/imagenette/2/00220.jpg',\n",
       " 'data/imagenette/2/00221.jpg',\n",
       " 'data/imagenette/2/00222.jpg',\n",
       " 'data/imagenette/2/00223.jpg',\n",
       " 'data/imagenette/2/00224.jpg',\n",
       " 'data/imagenette/2/00225.jpg',\n",
       " 'data/imagenette/2/00226.jpg',\n",
       " 'data/imagenette/2/00227.jpg',\n",
       " 'data/imagenette/2/00228.jpg',\n",
       " 'data/imagenette/2/00229.jpg',\n",
       " 'data/imagenette/2/00230.jpg',\n",
       " 'data/imagenette/2/00231.jpg',\n",
       " 'data/imagenette/2/00232.jpg',\n",
       " 'data/imagenette/2/00233.jpg',\n",
       " 'data/imagenette/2/00234.jpg',\n",
       " 'data/imagenette/2/00235.jpg',\n",
       " 'data/imagenette/2/00236.jpg',\n",
       " 'data/imagenette/2/00237.jpg',\n",
       " 'data/imagenette/2/00238.jpg',\n",
       " 'data/imagenette/2/00239.jpg',\n",
       " 'data/imagenette/2/00240.jpg',\n",
       " 'data/imagenette/2/00241.jpg',\n",
       " 'data/imagenette/2/00242.jpg',\n",
       " 'data/imagenette/2/00243.jpg',\n",
       " 'data/imagenette/2/00244.jpg',\n",
       " 'data/imagenette/2/00245.jpg',\n",
       " 'data/imagenette/2/00246.jpg',\n",
       " 'data/imagenette/2/00247.jpg',\n",
       " 'data/imagenette/2/00248.jpg',\n",
       " 'data/imagenette/2/00249.jpg',\n",
       " 'data/imagenette/2/00250.jpg',\n",
       " 'data/imagenette/2/00251.jpg',\n",
       " 'data/imagenette/2/00252.jpg',\n",
       " 'data/imagenette/2/00253.jpg',\n",
       " 'data/imagenette/2/00254.jpg',\n",
       " 'data/imagenette/2/00255.jpg',\n",
       " 'data/imagenette/2/00256.jpg',\n",
       " 'data/imagenette/2/00257.jpg',\n",
       " 'data/imagenette/2/00258.jpg',\n",
       " 'data/imagenette/2/00259.jpg',\n",
       " 'data/imagenette/2/00260.jpg',\n",
       " 'data/imagenette/2/00261.jpg',\n",
       " 'data/imagenette/2/00262.jpg',\n",
       " 'data/imagenette/2/00263.jpg',\n",
       " 'data/imagenette/2/00264.jpg',\n",
       " 'data/imagenette/2/00265.jpg',\n",
       " 'data/imagenette/2/00266.jpg',\n",
       " 'data/imagenette/2/00267.jpg',\n",
       " 'data/imagenette/2/00268.jpg',\n",
       " 'data/imagenette/2/00269.jpg',\n",
       " 'data/imagenette/2/00270.jpg',\n",
       " 'data/imagenette/2/00271.jpg',\n",
       " 'data/imagenette/2/00272.jpg',\n",
       " 'data/imagenette/2/00273.jpg',\n",
       " 'data/imagenette/2/00274.jpg',\n",
       " 'data/imagenette/2/00275.jpg',\n",
       " 'data/imagenette/2/00276.jpg',\n",
       " 'data/imagenette/2/00277.jpg',\n",
       " 'data/imagenette/2/00278.jpg',\n",
       " 'data/imagenette/2/00279.jpg',\n",
       " 'data/imagenette/2/00280.jpg',\n",
       " 'data/imagenette/2/00281.jpg',\n",
       " 'data/imagenette/2/00282.jpg',\n",
       " 'data/imagenette/2/00283.jpg',\n",
       " 'data/imagenette/2/00284.jpg',\n",
       " 'data/imagenette/2/00285.jpg',\n",
       " 'data/imagenette/2/00286.jpg',\n",
       " 'data/imagenette/2/00287.jpg',\n",
       " 'data/imagenette/2/00288.jpg',\n",
       " 'data/imagenette/2/00289.jpg',\n",
       " 'data/imagenette/2/00290.jpg',\n",
       " 'data/imagenette/2/00291.jpg',\n",
       " 'data/imagenette/2/00292.jpg',\n",
       " 'data/imagenette/2/00293.jpg',\n",
       " 'data/imagenette/2/00294.jpg',\n",
       " 'data/imagenette/2/00295.jpg',\n",
       " 'data/imagenette/2/00296.jpg',\n",
       " 'data/imagenette/2/00297.jpg',\n",
       " 'data/imagenette/2/00298.jpg',\n",
       " 'data/imagenette/2/00299.jpg',\n",
       " 'data/imagenette/2/00300.jpg',\n",
       " 'data/imagenette/2/00301.jpg',\n",
       " 'data/imagenette/2/00302.jpg',\n",
       " 'data/imagenette/2/00303.jpg',\n",
       " 'data/imagenette/2/00304.jpg',\n",
       " 'data/imagenette/2/00305.jpg',\n",
       " 'data/imagenette/2/00306.jpg',\n",
       " 'data/imagenette/2/00307.jpg',\n",
       " 'data/imagenette/2/00308.jpg',\n",
       " 'data/imagenette/2/00309.jpg',\n",
       " 'data/imagenette/2/00310.jpg',\n",
       " 'data/imagenette/2/00311.jpg',\n",
       " 'data/imagenette/2/00312.jpg',\n",
       " 'data/imagenette/2/00313.jpg',\n",
       " 'data/imagenette/2/00314.jpg',\n",
       " 'data/imagenette/2/00315.jpg',\n",
       " 'data/imagenette/2/00316.jpg',\n",
       " 'data/imagenette/2/00317.jpg',\n",
       " 'data/imagenette/2/00318.jpg',\n",
       " 'data/imagenette/2/00319.jpg',\n",
       " 'data/imagenette/2/00320.jpg',\n",
       " 'data/imagenette/2/00321.jpg',\n",
       " 'data/imagenette/2/00322.jpg',\n",
       " 'data/imagenette/2/00323.jpg',\n",
       " 'data/imagenette/2/00324.jpg',\n",
       " 'data/imagenette/2/00325.jpg',\n",
       " 'data/imagenette/2/00326.jpg',\n",
       " 'data/imagenette/2/00327.jpg',\n",
       " 'data/imagenette/2/00328.jpg',\n",
       " 'data/imagenette/2/00329.jpg',\n",
       " 'data/imagenette/2/00330.jpg',\n",
       " 'data/imagenette/2/00331.jpg',\n",
       " 'data/imagenette/2/00332.jpg',\n",
       " 'data/imagenette/2/00333.jpg',\n",
       " 'data/imagenette/2/00334.jpg',\n",
       " 'data/imagenette/2/00335.jpg',\n",
       " 'data/imagenette/2/00336.jpg',\n",
       " 'data/imagenette/2/00337.jpg',\n",
       " 'data/imagenette/2/00338.jpg',\n",
       " 'data/imagenette/2/00339.jpg',\n",
       " 'data/imagenette/2/00340.jpg',\n",
       " 'data/imagenette/2/00341.jpg',\n",
       " 'data/imagenette/2/00342.jpg',\n",
       " 'data/imagenette/2/00343.jpg',\n",
       " 'data/imagenette/2/00344.jpg',\n",
       " 'data/imagenette/2/00345.jpg',\n",
       " 'data/imagenette/2/00346.jpg',\n",
       " 'data/imagenette/2/00347.jpg',\n",
       " 'data/imagenette/2/00348.jpg',\n",
       " 'data/imagenette/2/00349.jpg',\n",
       " 'data/imagenette/2/00350.jpg',\n",
       " 'data/imagenette/2/00351.jpg',\n",
       " 'data/imagenette/2/00352.jpg',\n",
       " 'data/imagenette/2/00353.jpg',\n",
       " 'data/imagenette/2/00354.jpg',\n",
       " 'data/imagenette/2/00355.jpg',\n",
       " 'data/imagenette/2/00356.jpg',\n",
       " 'data/imagenette/2/00357.jpg',\n",
       " 'data/imagenette/2/00358.jpg',\n",
       " 'data/imagenette/2/00359.jpg',\n",
       " 'data/imagenette/2/00360.jpg',\n",
       " 'data/imagenette/2/00361.jpg',\n",
       " 'data/imagenette/2/00362.jpg',\n",
       " 'data/imagenette/2/00363.jpg',\n",
       " 'data/imagenette/2/00364.jpg',\n",
       " 'data/imagenette/2/00365.jpg',\n",
       " 'data/imagenette/2/00366.jpg',\n",
       " 'data/imagenette/2/00367.jpg',\n",
       " 'data/imagenette/2/00368.jpg',\n",
       " 'data/imagenette/2/00369.jpg',\n",
       " 'data/imagenette/2/00370.jpg',\n",
       " 'data/imagenette/2/00371.jpg',\n",
       " 'data/imagenette/2/00372.jpg',\n",
       " 'data/imagenette/2/00373.jpg',\n",
       " 'data/imagenette/2/00374.jpg',\n",
       " 'data/imagenette/2/00375.jpg',\n",
       " 'data/imagenette/2/00376.jpg',\n",
       " 'data/imagenette/2/00377.jpg',\n",
       " 'data/imagenette/2/00378.jpg',\n",
       " 'data/imagenette/2/00379.jpg',\n",
       " 'data/imagenette/2/00380.jpg',\n",
       " 'data/imagenette/2/00381.jpg',\n",
       " 'data/imagenette/2/00382.jpg',\n",
       " 'data/imagenette/2/00383.jpg',\n",
       " 'data/imagenette/2/00384.jpg',\n",
       " 'data/imagenette/2/00385.jpg',\n",
       " 'data/imagenette/2/00386.jpg',\n",
       " 'data/imagenette/2/00387.jpg',\n",
       " 'data/imagenette/2/00388.jpg',\n",
       " 'data/imagenette/2/00389.jpg',\n",
       " 'data/imagenette/2/00390.jpg',\n",
       " 'data/imagenette/2/00391.jpg',\n",
       " 'data/imagenette/2/00392.jpg',\n",
       " 'data/imagenette/2/00393.jpg',\n",
       " 'data/imagenette/2/00394.jpg',\n",
       " 'data/imagenette/2/00395.jpg',\n",
       " 'data/imagenette/2/00396.jpg',\n",
       " 'data/imagenette/2/00397.jpg',\n",
       " 'data/imagenette/2/00398.jpg',\n",
       " 'data/imagenette/2/00399.jpg',\n",
       " 'data/imagenette/2/00400.jpg',\n",
       " 'data/imagenette/2/00401.jpg',\n",
       " 'data/imagenette/2/00402.jpg',\n",
       " 'data/imagenette/2/00403.jpg',\n",
       " 'data/imagenette/2/00404.jpg',\n",
       " 'data/imagenette/2/00405.jpg',\n",
       " 'data/imagenette/2/00406.jpg',\n",
       " 'data/imagenette/2/00407.jpg',\n",
       " 'data/imagenette/2/00408.jpg',\n",
       " 'data/imagenette/2/00409.jpg',\n",
       " 'data/imagenette/2/00410.jpg',\n",
       " 'data/imagenette/2/00411.jpg',\n",
       " 'data/imagenette/2/00412.jpg',\n",
       " 'data/imagenette/2/00413.jpg',\n",
       " 'data/imagenette/2/00414.jpg',\n",
       " 'data/imagenette/2/00415.jpg',\n",
       " 'data/imagenette/2/00416.jpg',\n",
       " 'data/imagenette/2/00417.jpg',\n",
       " 'data/imagenette/2/00418.jpg',\n",
       " 'data/imagenette/2/00419.jpg',\n",
       " 'data/imagenette/2/00421.jpg',\n",
       " 'data/imagenette/2/00422.jpg',\n",
       " 'data/imagenette/2/00423.jpg',\n",
       " 'data/imagenette/2/00424.jpg',\n",
       " 'data/imagenette/2/00425.jpg',\n",
       " 'data/imagenette/2/00426.jpg',\n",
       " 'data/imagenette/2/00427.jpg',\n",
       " 'data/imagenette/2/00428.jpg',\n",
       " 'data/imagenette/2/00429.jpg',\n",
       " 'data/imagenette/2/00430.jpg',\n",
       " 'data/imagenette/2/00431.jpg',\n",
       " 'data/imagenette/2/00432.jpg',\n",
       " 'data/imagenette/2/00433.jpg',\n",
       " 'data/imagenette/2/00434.jpg',\n",
       " 'data/imagenette/2/00435.jpg',\n",
       " 'data/imagenette/2/00436.jpg',\n",
       " 'data/imagenette/2/00437.jpg',\n",
       " 'data/imagenette/2/00438.jpg',\n",
       " 'data/imagenette/2/00439.jpg',\n",
       " 'data/imagenette/2/00440.jpg',\n",
       " 'data/imagenette/2/00441.jpg',\n",
       " 'data/imagenette/2/00442.jpg',\n",
       " 'data/imagenette/2/00443.jpg',\n",
       " 'data/imagenette/2/00444.jpg',\n",
       " 'data/imagenette/2/00445.jpg',\n",
       " 'data/imagenette/2/00446.jpg',\n",
       " 'data/imagenette/2/00447.jpg',\n",
       " 'data/imagenette/2/00448.jpg',\n",
       " 'data/imagenette/2/00449.jpg',\n",
       " 'data/imagenette/2/00450.jpg',\n",
       " 'data/imagenette/2/00451.jpg',\n",
       " 'data/imagenette/2/00452.jpg',\n",
       " 'data/imagenette/2/00453.jpg',\n",
       " 'data/imagenette/2/00454.jpg',\n",
       " 'data/imagenette/2/00455.jpg',\n",
       " 'data/imagenette/2/00456.jpg',\n",
       " 'data/imagenette/2/00457.jpg',\n",
       " 'data/imagenette/2/00458.jpg',\n",
       " 'data/imagenette/2/00459.jpg',\n",
       " 'data/imagenette/2/00460.jpg',\n",
       " 'data/imagenette/2/00461.jpg',\n",
       " 'data/imagenette/2/00462.jpg',\n",
       " 'data/imagenette/2/00463.jpg',\n",
       " 'data/imagenette/2/00464.jpg',\n",
       " 'data/imagenette/2/00465.jpg',\n",
       " 'data/imagenette/2/00466.jpg',\n",
       " 'data/imagenette/2/00467.jpg',\n",
       " 'data/imagenette/2/00468.jpg',\n",
       " 'data/imagenette/2/00469.jpg',\n",
       " 'data/imagenette/2/00470.jpg',\n",
       " 'data/imagenette/2/00471.jpg',\n",
       " 'data/imagenette/2/00472.jpg',\n",
       " 'data/imagenette/2/00473.jpg',\n",
       " 'data/imagenette/2/00474.jpg',\n",
       " 'data/imagenette/2/00475.jpg',\n",
       " 'data/imagenette/2/00476.jpg',\n",
       " 'data/imagenette/2/00477.jpg',\n",
       " 'data/imagenette/2/00478.jpg',\n",
       " 'data/imagenette/2/00479.jpg',\n",
       " 'data/imagenette/2/00480.jpg',\n",
       " 'data/imagenette/2/00481.jpg',\n",
       " 'data/imagenette/2/00482.jpg',\n",
       " 'data/imagenette/2/00483.jpg',\n",
       " 'data/imagenette/2/00484.jpg',\n",
       " 'data/imagenette/2/00485.jpg',\n",
       " 'data/imagenette/2/00486.jpg',\n",
       " 'data/imagenette/2/00487.jpg',\n",
       " 'data/imagenette/2/00488.jpg',\n",
       " 'data/imagenette/2/00489.jpg',\n",
       " 'data/imagenette/2/00490.jpg',\n",
       " 'data/imagenette/2/00491.jpg',\n",
       " 'data/imagenette/2/00492.jpg',\n",
       " 'data/imagenette/2/00493.jpg',\n",
       " 'data/imagenette/2/00494.jpg',\n",
       " 'data/imagenette/2/00495.jpg',\n",
       " 'data/imagenette/2/00496.jpg',\n",
       " 'data/imagenette/2/00497.jpg',\n",
       " 'data/imagenette/2/00498.jpg',\n",
       " 'data/imagenette/2/00499.jpg',\n",
       " 'data/imagenette/2/00500.jpg',\n",
       " 'data/imagenette/2/00501.jpg',\n",
       " 'data/imagenette/2/00502.jpg',\n",
       " 'data/imagenette/2/00503.jpg',\n",
       " 'data/imagenette/2/00504.jpg',\n",
       " 'data/imagenette/2/00505.jpg',\n",
       " 'data/imagenette/2/00507.jpg',\n",
       " 'data/imagenette/2/00508.jpg',\n",
       " 'data/imagenette/2/00509.jpg',\n",
       " 'data/imagenette/2/00510.jpg',\n",
       " 'data/imagenette/2/00511.jpg',\n",
       " 'data/imagenette/2/00512.jpg',\n",
       " 'data/imagenette/2/00513.jpg',\n",
       " 'data/imagenette/2/00514.jpg',\n",
       " 'data/imagenette/2/00515.jpg',\n",
       " 'data/imagenette/2/00516.jpg',\n",
       " 'data/imagenette/2/00517.jpg',\n",
       " 'data/imagenette/2/00518.jpg',\n",
       " 'data/imagenette/2/00519.jpg',\n",
       " 'data/imagenette/2/00520.jpg',\n",
       " 'data/imagenette/2/00521.jpg',\n",
       " 'data/imagenette/2/00522.jpg',\n",
       " 'data/imagenette/2/00523.jpg',\n",
       " 'data/imagenette/2/00524.jpg',\n",
       " 'data/imagenette/2/00525.jpg',\n",
       " 'data/imagenette/2/00526.jpg',\n",
       " 'data/imagenette/2/00527.jpg',\n",
       " 'data/imagenette/2/00528.jpg',\n",
       " 'data/imagenette/2/00529.jpg',\n",
       " 'data/imagenette/2/00530.jpg',\n",
       " 'data/imagenette/2/00531.jpg',\n",
       " 'data/imagenette/2/00532.jpg',\n",
       " 'data/imagenette/2/00533.jpg',\n",
       " 'data/imagenette/2/00534.jpg',\n",
       " 'data/imagenette/2/00535.jpg',\n",
       " 'data/imagenette/2/00536.jpg',\n",
       " 'data/imagenette/2/00537.jpg',\n",
       " 'data/imagenette/2/00538.jpg',\n",
       " 'data/imagenette/2/00539.jpg',\n",
       " 'data/imagenette/2/00540.jpg',\n",
       " 'data/imagenette/2/00541.jpg',\n",
       " 'data/imagenette/2/00542.jpg',\n",
       " 'data/imagenette/2/00543.jpg',\n",
       " 'data/imagenette/2/00544.jpg',\n",
       " 'data/imagenette/2/00545.jpg',\n",
       " 'data/imagenette/2/00546.jpg',\n",
       " 'data/imagenette/2/00547.jpg',\n",
       " 'data/imagenette/2/00548.jpg',\n",
       " 'data/imagenette/2/00549.jpg',\n",
       " 'data/imagenette/2/00550.jpg',\n",
       " 'data/imagenette/2/00551.jpg',\n",
       " 'data/imagenette/2/00552.jpg',\n",
       " 'data/imagenette/2/00553.jpg',\n",
       " 'data/imagenette/2/00554.jpg',\n",
       " 'data/imagenette/2/00555.jpg',\n",
       " 'data/imagenette/2/00556.jpg',\n",
       " 'data/imagenette/2/00557.jpg',\n",
       " 'data/imagenette/2/00558.jpg',\n",
       " 'data/imagenette/2/00559.jpg',\n",
       " 'data/imagenette/2/00560.jpg',\n",
       " 'data/imagenette/2/00561.jpg',\n",
       " 'data/imagenette/2/00562.jpg',\n",
       " 'data/imagenette/2/00563.jpg',\n",
       " 'data/imagenette/2/00564.jpg',\n",
       " 'data/imagenette/2/00565.jpg',\n",
       " 'data/imagenette/2/00566.jpg',\n",
       " 'data/imagenette/2/00567.jpg',\n",
       " 'data/imagenette/2/00568.jpg',\n",
       " 'data/imagenette/2/00569.jpg',\n",
       " 'data/imagenette/2/00570.jpg',\n",
       " 'data/imagenette/2/00571.jpg',\n",
       " 'data/imagenette/2/00572.jpg',\n",
       " 'data/imagenette/2/00573.jpg',\n",
       " 'data/imagenette/2/00574.jpg',\n",
       " 'data/imagenette/2/00576.jpg',\n",
       " 'data/imagenette/2/00577.jpg',\n",
       " 'data/imagenette/2/00578.jpg',\n",
       " 'data/imagenette/2/00579.jpg',\n",
       " 'data/imagenette/2/00580.jpg',\n",
       " 'data/imagenette/2/00581.jpg',\n",
       " 'data/imagenette/2/00582.jpg',\n",
       " 'data/imagenette/2/00583.jpg',\n",
       " 'data/imagenette/2/00584.jpg',\n",
       " 'data/imagenette/2/00585.jpg',\n",
       " 'data/imagenette/2/00586.jpg',\n",
       " 'data/imagenette/2/00587.jpg',\n",
       " 'data/imagenette/2/00588.jpg',\n",
       " 'data/imagenette/2/00589.jpg',\n",
       " 'data/imagenette/2/00590.jpg',\n",
       " 'data/imagenette/2/00591.jpg',\n",
       " 'data/imagenette/2/00592.jpg',\n",
       " 'data/imagenette/2/00593.jpg',\n",
       " 'data/imagenette/2/00594.jpg',\n",
       " 'data/imagenette/2/00595.jpg',\n",
       " 'data/imagenette/2/00596.jpg',\n",
       " 'data/imagenette/2/00597.jpg',\n",
       " 'data/imagenette/2/00598.jpg',\n",
       " 'data/imagenette/2/00599.jpg',\n",
       " 'data/imagenette/2/00600.jpg',\n",
       " 'data/imagenette/2/00601.jpg',\n",
       " 'data/imagenette/2/00602.jpg',\n",
       " 'data/imagenette/2/00603.jpg',\n",
       " 'data/imagenette/2/00604.jpg',\n",
       " 'data/imagenette/2/00605.jpg',\n",
       " 'data/imagenette/2/00606.jpg',\n",
       " 'data/imagenette/2/00607.jpg',\n",
       " 'data/imagenette/2/00608.jpg',\n",
       " 'data/imagenette/2/00609.jpg',\n",
       " 'data/imagenette/2/00610.jpg',\n",
       " 'data/imagenette/2/00611.jpg',\n",
       " 'data/imagenette/2/00612.jpg',\n",
       " 'data/imagenette/2/00613.jpg',\n",
       " 'data/imagenette/2/00614.jpg',\n",
       " 'data/imagenette/2/00615.jpg',\n",
       " 'data/imagenette/2/00616.jpg',\n",
       " 'data/imagenette/2/00617.jpg',\n",
       " 'data/imagenette/2/00618.jpg',\n",
       " 'data/imagenette/2/00619.jpg',\n",
       " 'data/imagenette/2/00620.jpg',\n",
       " 'data/imagenette/2/00621.jpg',\n",
       " 'data/imagenette/2/00622.jpg',\n",
       " 'data/imagenette/2/00623.jpg',\n",
       " 'data/imagenette/2/00624.jpg',\n",
       " 'data/imagenette/2/00625.jpg',\n",
       " 'data/imagenette/2/00626.jpg',\n",
       " 'data/imagenette/2/00627.jpg',\n",
       " 'data/imagenette/2/00628.jpg',\n",
       " 'data/imagenette/2/00629.jpg',\n",
       " 'data/imagenette/2/00630.jpg',\n",
       " 'data/imagenette/2/00631.jpg',\n",
       " 'data/imagenette/2/00632.jpg',\n",
       " 'data/imagenette/2/00633.jpg',\n",
       " 'data/imagenette/2/00634.jpg',\n",
       " 'data/imagenette/2/00635.jpg',\n",
       " 'data/imagenette/2/00636.jpg',\n",
       " 'data/imagenette/2/00637.jpg',\n",
       " 'data/imagenette/2/00638.jpg',\n",
       " 'data/imagenette/2/00639.jpg',\n",
       " 'data/imagenette/2/00640.jpg',\n",
       " 'data/imagenette/2/00641.jpg',\n",
       " 'data/imagenette/2/00642.jpg',\n",
       " 'data/imagenette/2/00643.jpg',\n",
       " 'data/imagenette/2/00644.jpg',\n",
       " 'data/imagenette/2/00645.jpg',\n",
       " 'data/imagenette/2/00646.jpg',\n",
       " 'data/imagenette/2/00647.jpg',\n",
       " 'data/imagenette/2/00648.jpg',\n",
       " 'data/imagenette/2/00649.jpg',\n",
       " 'data/imagenette/2/00650.jpg',\n",
       " 'data/imagenette/2/00651.jpg',\n",
       " 'data/imagenette/2/00652.jpg',\n",
       " 'data/imagenette/2/00653.jpg',\n",
       " 'data/imagenette/2/00654.jpg',\n",
       " 'data/imagenette/2/00655.jpg',\n",
       " 'data/imagenette/2/00656.jpg',\n",
       " 'data/imagenette/2/00657.jpg',\n",
       " 'data/imagenette/2/00658.jpg',\n",
       " 'data/imagenette/2/00659.jpg',\n",
       " 'data/imagenette/2/00660.jpg',\n",
       " 'data/imagenette/2/00661.jpg',\n",
       " 'data/imagenette/2/00662.jpg',\n",
       " 'data/imagenette/2/00663.jpg',\n",
       " 'data/imagenette/2/00664.jpg',\n",
       " 'data/imagenette/2/00665.jpg',\n",
       " 'data/imagenette/2/00666.jpg',\n",
       " 'data/imagenette/2/00667.jpg',\n",
       " 'data/imagenette/2/00668.jpg',\n",
       " 'data/imagenette/2/00669.jpg',\n",
       " 'data/imagenette/2/00670.jpg',\n",
       " 'data/imagenette/2/00671.jpg',\n",
       " 'data/imagenette/2/00672.jpg',\n",
       " 'data/imagenette/2/00673.jpg',\n",
       " 'data/imagenette/2/00674.jpg',\n",
       " 'data/imagenette/2/00675.jpg',\n",
       " 'data/imagenette/2/00676.jpg',\n",
       " 'data/imagenette/2/00677.jpg',\n",
       " 'data/imagenette/2/00678.jpg',\n",
       " 'data/imagenette/2/00679.jpg',\n",
       " 'data/imagenette/2/00680.jpg',\n",
       " 'data/imagenette/2/00681.jpg',\n",
       " 'data/imagenette/2/00682.jpg',\n",
       " 'data/imagenette/2/00683.jpg',\n",
       " 'data/imagenette/2/00684.jpg',\n",
       " 'data/imagenette/2/00685.jpg',\n",
       " 'data/imagenette/2/00686.jpg',\n",
       " 'data/imagenette/2/00687.jpg',\n",
       " 'data/imagenette/2/00688.jpg',\n",
       " 'data/imagenette/2/00689.jpg',\n",
       " 'data/imagenette/2/00690.jpg',\n",
       " 'data/imagenette/2/00691.jpg',\n",
       " 'data/imagenette/2/00692.jpg',\n",
       " 'data/imagenette/2/00693.jpg',\n",
       " 'data/imagenette/2/00694.jpg',\n",
       " 'data/imagenette/2/00695.jpg',\n",
       " 'data/imagenette/2/00696.jpg',\n",
       " 'data/imagenette/2/00697.jpg',\n",
       " 'data/imagenette/2/00698.jpg',\n",
       " 'data/imagenette/2/00699.jpg',\n",
       " 'data/imagenette/2/00700.jpg',\n",
       " 'data/imagenette/2/00701.jpg',\n",
       " 'data/imagenette/2/00702.jpg',\n",
       " 'data/imagenette/2/00703.jpg',\n",
       " 'data/imagenette/2/00704.jpg',\n",
       " 'data/imagenette/2/00705.jpg',\n",
       " 'data/imagenette/2/00706.jpg',\n",
       " 'data/imagenette/2/00707.jpg',\n",
       " 'data/imagenette/2/00708.jpg',\n",
       " 'data/imagenette/2/00709.jpg',\n",
       " 'data/imagenette/2/00711.jpg',\n",
       " 'data/imagenette/2/00712.jpg',\n",
       " 'data/imagenette/2/00713.jpg',\n",
       " 'data/imagenette/2/00714.jpg',\n",
       " 'data/imagenette/2/00715.jpg',\n",
       " 'data/imagenette/2/00716.jpg',\n",
       " 'data/imagenette/2/00717.jpg',\n",
       " 'data/imagenette/2/00718.jpg',\n",
       " 'data/imagenette/2/00719.jpg',\n",
       " 'data/imagenette/2/00720.jpg',\n",
       " 'data/imagenette/2/00721.jpg',\n",
       " 'data/imagenette/2/00722.jpg',\n",
       " 'data/imagenette/2/00723.jpg',\n",
       " 'data/imagenette/2/00724.jpg',\n",
       " 'data/imagenette/2/00725.jpg',\n",
       " 'data/imagenette/2/00726.jpg',\n",
       " 'data/imagenette/2/00727.jpg',\n",
       " 'data/imagenette/2/00728.jpg',\n",
       " 'data/imagenette/2/00729.jpg',\n",
       " 'data/imagenette/2/00730.jpg',\n",
       " 'data/imagenette/2/00731.jpg',\n",
       " 'data/imagenette/2/00732.jpg',\n",
       " 'data/imagenette/2/00733.jpg',\n",
       " 'data/imagenette/2/00734.jpg',\n",
       " 'data/imagenette/2/00735.jpg',\n",
       " 'data/imagenette/2/00736.jpg',\n",
       " 'data/imagenette/2/00737.jpg',\n",
       " 'data/imagenette/2/00738.jpg',\n",
       " 'data/imagenette/2/00739.jpg',\n",
       " 'data/imagenette/2/00740.jpg',\n",
       " 'data/imagenette/2/00741.jpg',\n",
       " 'data/imagenette/2/00742.jpg',\n",
       " 'data/imagenette/2/00743.jpg',\n",
       " 'data/imagenette/2/00744.jpg',\n",
       " 'data/imagenette/2/00745.jpg',\n",
       " 'data/imagenette/2/00746.jpg',\n",
       " 'data/imagenette/2/00747.jpg',\n",
       " 'data/imagenette/2/00748.jpg',\n",
       " 'data/imagenette/2/00749.jpg',\n",
       " 'data/imagenette/2/00750.jpg',\n",
       " 'data/imagenette/2/00751.jpg',\n",
       " 'data/imagenette/2/00752.jpg',\n",
       " 'data/imagenette/2/00753.jpg',\n",
       " 'data/imagenette/2/00754.jpg',\n",
       " 'data/imagenette/2/00755.jpg',\n",
       " 'data/imagenette/2/00756.jpg',\n",
       " 'data/imagenette/2/00757.jpg',\n",
       " 'data/imagenette/2/00758.jpg',\n",
       " 'data/imagenette/2/00759.jpg',\n",
       " 'data/imagenette/2/00760.jpg',\n",
       " 'data/imagenette/2/00761.jpg',\n",
       " 'data/imagenette/2/00762.jpg',\n",
       " 'data/imagenette/2/00763.jpg',\n",
       " 'data/imagenette/2/00764.jpg',\n",
       " 'data/imagenette/2/00765.jpg',\n",
       " 'data/imagenette/2/00766.jpg',\n",
       " 'data/imagenette/2/00767.jpg',\n",
       " 'data/imagenette/2/00768.jpg',\n",
       " 'data/imagenette/2/00769.jpg',\n",
       " 'data/imagenette/2/00770.jpg',\n",
       " 'data/imagenette/2/00771.jpg',\n",
       " 'data/imagenette/2/00772.jpg',\n",
       " 'data/imagenette/2/00773.jpg',\n",
       " 'data/imagenette/2/00774.jpg',\n",
       " 'data/imagenette/2/00775.jpg',\n",
       " 'data/imagenette/2/00776.jpg',\n",
       " 'data/imagenette/2/00777.jpg',\n",
       " 'data/imagenette/2/00778.jpg',\n",
       " 'data/imagenette/2/00779.jpg',\n",
       " 'data/imagenette/2/00780.jpg',\n",
       " 'data/imagenette/2/00781.jpg',\n",
       " 'data/imagenette/2/00782.jpg',\n",
       " 'data/imagenette/2/00783.jpg',\n",
       " 'data/imagenette/2/00784.jpg',\n",
       " 'data/imagenette/2/00785.jpg',\n",
       " 'data/imagenette/2/00786.jpg',\n",
       " 'data/imagenette/2/00787.jpg',\n",
       " 'data/imagenette/2/00788.jpg',\n",
       " 'data/imagenette/2/00789.jpg',\n",
       " 'data/imagenette/2/00790.jpg',\n",
       " 'data/imagenette/2/00791.jpg',\n",
       " 'data/imagenette/2/00792.jpg',\n",
       " 'data/imagenette/2/00793.jpg',\n",
       " 'data/imagenette/2/00794.jpg',\n",
       " 'data/imagenette/2/00795.jpg',\n",
       " 'data/imagenette/2/00796.jpg',\n",
       " 'data/imagenette/2/00797.jpg',\n",
       " 'data/imagenette/2/00798.jpg',\n",
       " 'data/imagenette/2/00799.jpg',\n",
       " 'data/imagenette/2/00800.jpg',\n",
       " 'data/imagenette/2/00801.jpg',\n",
       " 'data/imagenette/2/00802.jpg',\n",
       " 'data/imagenette/2/00803.jpg',\n",
       " 'data/imagenette/2/00804.jpg',\n",
       " 'data/imagenette/2/00805.jpg',\n",
       " 'data/imagenette/2/00806.jpg',\n",
       " 'data/imagenette/2/00807.jpg',\n",
       " 'data/imagenette/2/00808.jpg',\n",
       " 'data/imagenette/2/00809.jpg',\n",
       " 'data/imagenette/2/00810.jpg',\n",
       " 'data/imagenette/2/00811.jpg',\n",
       " 'data/imagenette/2/00812.jpg',\n",
       " 'data/imagenette/2/00813.jpg',\n",
       " 'data/imagenette/2/00814.jpg',\n",
       " 'data/imagenette/2/00815.jpg',\n",
       " 'data/imagenette/2/00816.jpg',\n",
       " 'data/imagenette/2/00817.jpg',\n",
       " 'data/imagenette/2/00818.jpg',\n",
       " 'data/imagenette/2/00819.jpg',\n",
       " 'data/imagenette/2/00820.jpg',\n",
       " 'data/imagenette/2/00821.jpg',\n",
       " 'data/imagenette/2/00822.jpg',\n",
       " 'data/imagenette/2/00823.jpg',\n",
       " 'data/imagenette/2/00824.jpg',\n",
       " 'data/imagenette/2/00825.jpg',\n",
       " 'data/imagenette/2/00826.jpg',\n",
       " 'data/imagenette/2/00827.jpg',\n",
       " 'data/imagenette/2/00828.jpg',\n",
       " 'data/imagenette/2/00829.jpg',\n",
       " 'data/imagenette/2/00830.jpg',\n",
       " 'data/imagenette/2/00831.jpg',\n",
       " 'data/imagenette/2/00832.jpg',\n",
       " 'data/imagenette/2/00833.jpg',\n",
       " 'data/imagenette/2/00834.jpg',\n",
       " 'data/imagenette/2/00835.jpg',\n",
       " 'data/imagenette/2/00836.jpg',\n",
       " 'data/imagenette/2/00837.jpg',\n",
       " 'data/imagenette/2/00838.jpg',\n",
       " 'data/imagenette/2/00839.jpg',\n",
       " 'data/imagenette/2/00840.jpg',\n",
       " 'data/imagenette/2/00841.jpg',\n",
       " 'data/imagenette/2/00842.jpg',\n",
       " 'data/imagenette/2/00843.jpg',\n",
       " 'data/imagenette/2/00844.jpg',\n",
       " 'data/imagenette/2/00845.jpg',\n",
       " 'data/imagenette/2/00846.jpg',\n",
       " 'data/imagenette/2/00847.jpg',\n",
       " 'data/imagenette/2/00848.jpg',\n",
       " 'data/imagenette/2/00849.jpg',\n",
       " 'data/imagenette/2/00850.jpg',\n",
       " 'data/imagenette/2/00851.jpg',\n",
       " 'data/imagenette/2/00852.jpg',\n",
       " 'data/imagenette/2/00853.jpg',\n",
       " 'data/imagenette/2/00854.jpg',\n",
       " 'data/imagenette/2/00855.jpg',\n",
       " 'data/imagenette/2/00856.jpg',\n",
       " 'data/imagenette/2/00857.jpg',\n",
       " 'data/imagenette/2/00858.jpg',\n",
       " 'data/imagenette/2/00859.jpg',\n",
       " 'data/imagenette/2/00860.jpg',\n",
       " 'data/imagenette/2/00861.jpg',\n",
       " 'data/imagenette/2/00862.jpg',\n",
       " 'data/imagenette/2/00863.jpg',\n",
       " 'data/imagenette/2/00864.jpg',\n",
       " 'data/imagenette/2/00865.jpg',\n",
       " 'data/imagenette/2/00866.jpg',\n",
       " 'data/imagenette/2/00867.jpg',\n",
       " 'data/imagenette/2/00868.jpg',\n",
       " 'data/imagenette/2/00869.jpg',\n",
       " 'data/imagenette/2/00870.jpg',\n",
       " 'data/imagenette/2/00871.jpg',\n",
       " 'data/imagenette/2/00872.jpg',\n",
       " 'data/imagenette/2/00873.jpg',\n",
       " 'data/imagenette/2/00874.jpg',\n",
       " 'data/imagenette/2/00875.jpg',\n",
       " 'data/imagenette/2/00876.jpg',\n",
       " 'data/imagenette/2/00877.jpg',\n",
       " 'data/imagenette/2/00878.jpg',\n",
       " 'data/imagenette/2/00879.jpg',\n",
       " 'data/imagenette/2/00880.jpg',\n",
       " 'data/imagenette/2/00881.jpg',\n",
       " 'data/imagenette/2/00882.jpg',\n",
       " 'data/imagenette/2/00883.jpg',\n",
       " 'data/imagenette/2/00884.jpg',\n",
       " 'data/imagenette/2/00885.jpg',\n",
       " 'data/imagenette/2/00886.jpg',\n",
       " 'data/imagenette/2/00887.jpg',\n",
       " 'data/imagenette/2/00888.jpg',\n",
       " 'data/imagenette/2/00889.jpg',\n",
       " 'data/imagenette/2/00890.jpg',\n",
       " 'data/imagenette/2/00891.jpg',\n",
       " 'data/imagenette/2/00892.jpg',\n",
       " 'data/imagenette/2/00893.jpg',\n",
       " 'data/imagenette/2/00894.jpg',\n",
       " 'data/imagenette/2/00895.jpg',\n",
       " 'data/imagenette/2/00896.jpg',\n",
       " 'data/imagenette/2/00897.jpg',\n",
       " 'data/imagenette/2/00898.jpg',\n",
       " 'data/imagenette/2/00899.jpg',\n",
       " 'data/imagenette/2/00900.jpg',\n",
       " 'data/imagenette/2/00901.jpg',\n",
       " 'data/imagenette/2/00902.jpg',\n",
       " 'data/imagenette/2/00903.jpg',\n",
       " 'data/imagenette/2/00904.jpg',\n",
       " 'data/imagenette/2/00905.jpg',\n",
       " 'data/imagenette/2/00906.jpg',\n",
       " 'data/imagenette/2/00907.jpg',\n",
       " 'data/imagenette/2/00908.jpg',\n",
       " 'data/imagenette/2/00909.jpg',\n",
       " 'data/imagenette/2/00910.jpg',\n",
       " 'data/imagenette/2/00911.jpg',\n",
       " 'data/imagenette/2/00912.jpg',\n",
       " 'data/imagenette/2/00913.jpg',\n",
       " 'data/imagenette/2/00914.jpg',\n",
       " 'data/imagenette/2/00915.jpg',\n",
       " 'data/imagenette/2/00916.jpg',\n",
       " 'data/imagenette/2/00917.jpg',\n",
       " 'data/imagenette/2/00918.jpg',\n",
       " 'data/imagenette/2/00919.jpg',\n",
       " 'data/imagenette/2/00920.jpg',\n",
       " 'data/imagenette/2/00921.jpg',\n",
       " 'data/imagenette/2/00922.jpg',\n",
       " 'data/imagenette/2/00923.jpg',\n",
       " 'data/imagenette/2/00924.jpg',\n",
       " 'data/imagenette/2/00925.jpg',\n",
       " 'data/imagenette/2/00926.jpg',\n",
       " 'data/imagenette/2/00927.jpg',\n",
       " 'data/imagenette/2/00928.jpg',\n",
       " 'data/imagenette/2/00929.jpg',\n",
       " 'data/imagenette/2/00930.jpg',\n",
       " 'data/imagenette/2/00931.jpg',\n",
       " 'data/imagenette/2/00932.jpg',\n",
       " 'data/imagenette/2/00933.jpg',\n",
       " 'data/imagenette/2/00934.jpg',\n",
       " 'data/imagenette/2/00935.jpg',\n",
       " 'data/imagenette/2/00936.jpg',\n",
       " 'data/imagenette/2/00937.jpg',\n",
       " 'data/imagenette/2/00938.jpg',\n",
       " 'data/imagenette/2/00939.jpg',\n",
       " 'data/imagenette/2/00940.jpg',\n",
       " 'data/imagenette/2/00941.jpg',\n",
       " 'data/imagenette/2/00942.jpg',\n",
       " 'data/imagenette/2/00943.jpg',\n",
       " 'data/imagenette/2/00944.jpg',\n",
       " 'data/imagenette/2/00945.jpg',\n",
       " 'data/imagenette/2/00946.jpg',\n",
       " 'data/imagenette/2/00947.jpg',\n",
       " 'data/imagenette/2/00948.jpg',\n",
       " 'data/imagenette/2/00949.jpg',\n",
       " 'data/imagenette/2/00950.jpg',\n",
       " 'data/imagenette/2/00951.jpg',\n",
       " 'data/imagenette/2/00952.jpg',\n",
       " 'data/imagenette/2/00953.jpg',\n",
       " 'data/imagenette/2/00954.jpg',\n",
       " 'data/imagenette/2/00955.jpg',\n",
       " 'data/imagenette/2/00956.jpg',\n",
       " 'data/imagenette/2/00957.jpg',\n",
       " 'data/imagenette/2/00958.jpg',\n",
       " 'data/imagenette/2/00959.jpg',\n",
       " 'data/imagenette/2/00960.jpg',\n",
       " 'data/imagenette/2/00961.jpg',\n",
       " 'data/imagenette/2/00962.jpg',\n",
       " 'data/imagenette/2/00963.jpg',\n",
       " 'data/imagenette/2/00964.jpg',\n",
       " 'data/imagenette/2/00965.jpg',\n",
       " 'data/imagenette/2/00966.jpg',\n",
       " 'data/imagenette/2/00967.jpg',\n",
       " 'data/imagenette/2/00968.jpg',\n",
       " 'data/imagenette/2/00969.jpg',\n",
       " 'data/imagenette/2/00970.jpg',\n",
       " 'data/imagenette/2/00971.jpg',\n",
       " 'data/imagenette/2/00972.jpg',\n",
       " 'data/imagenette/2/00973.jpg',\n",
       " 'data/imagenette/2/00975.jpg',\n",
       " 'data/imagenette/2/00976.jpg',\n",
       " 'data/imagenette/2/00977.jpg',\n",
       " 'data/imagenette/2/00978.jpg',\n",
       " 'data/imagenette/2/00979.jpg',\n",
       " 'data/imagenette/2/00980.jpg',\n",
       " 'data/imagenette/2/00981.jpg',\n",
       " 'data/imagenette/2/00982.jpg',\n",
       " 'data/imagenette/2/00983.jpg',\n",
       " 'data/imagenette/2/00984.jpg',\n",
       " 'data/imagenette/2/00985.jpg',\n",
       " 'data/imagenette/2/00986.jpg',\n",
       " 'data/imagenette/2/00987.jpg',\n",
       " 'data/imagenette/2/00988.jpg',\n",
       " 'data/imagenette/2/00989.jpg',\n",
       " 'data/imagenette/2/00990.jpg',\n",
       " 'data/imagenette/2/00991.jpg',\n",
       " 'data/imagenette/2/00992.jpg',\n",
       " 'data/imagenette/0/00993.jpg',\n",
       " 'data/imagenette/0/00994.jpg',\n",
       " 'data/imagenette/0/00995.jpg',\n",
       " 'data/imagenette/0/00996.jpg',\n",
       " 'data/imagenette/0/00997.jpg',\n",
       " 'data/imagenette/0/00998.jpg',\n",
       " 'data/imagenette/0/00999.jpg',\n",
       " 'data/imagenette/0/01000.jpg',\n",
       " 'data/imagenette/0/01001.jpg',\n",
       " 'data/imagenette/0/01002.jpg',\n",
       " 'data/imagenette/0/01003.jpg',\n",
       " 'data/imagenette/0/01004.jpg',\n",
       " 'data/imagenette/0/01005.jpg',\n",
       " 'data/imagenette/0/01006.jpg',\n",
       " 'data/imagenette/0/01007.jpg',\n",
       " 'data/imagenette/0/01008.jpg',\n",
       " 'data/imagenette/0/01009.jpg',\n",
       " ...]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_parquet(\"unlabeled_samples.parquet\")\n",
    "filepaths = df[\"filepath\"].tolist()\n",
    "\n",
    "filepaths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9369"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(filepaths)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:30:02.962\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36mpredict\u001b[0m:\u001b[36m77\u001b[0m - \u001b[1mRunning inference on 9369 samples\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "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>filepath</th>\n",
       "      <th>pred_label</th>\n",
       "      <th>pred_conf</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/imagenette/2/00000.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.220948</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/imagenette/2/00001.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.227241</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/imagenette/2/00002.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.208374</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>data/imagenette/2/00004.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.228580</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>data/imagenette/2/00005.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.134166</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9364</th>\n",
       "      <td>data/imagenette/5/09464.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.219891</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9365</th>\n",
       "      <td>data/imagenette/5/09465.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.231822</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9366</th>\n",
       "      <td>data/imagenette/5/09466.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.231926</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9367</th>\n",
       "      <td>data/imagenette/5/09467.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.226841</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9368</th>\n",
       "      <td>data/imagenette/5/09468.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.231906</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>9369 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                         filepath       pred_label  pred_conf\n",
       "0     data/imagenette/2/00000.jpg  cassette player   0.220948\n",
       "1     data/imagenette/2/00001.jpg  cassette player   0.227241\n",
       "2     data/imagenette/2/00002.jpg  cassette player   0.208374\n",
       "3     data/imagenette/2/00004.jpg  cassette player   0.228580\n",
       "4     data/imagenette/2/00005.jpg  cassette player   0.134166\n",
       "...                           ...              ...        ...\n",
       "9364  data/imagenette/5/09464.jpg      French horn   0.219891\n",
       "9365  data/imagenette/5/09465.jpg      French horn   0.231822\n",
       "9366  data/imagenette/5/09466.jpg      French horn   0.231926\n",
       "9367  data/imagenette/5/09467.jpg      French horn   0.226841\n",
       "9368  data/imagenette/5/09468.jpg      French horn   0.231906\n",
       "\n",
       "[9369 rows x 3 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_df = al.predict(filepaths, batch_size=128)\n",
    "pred_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2025-01-13 15:30:11.552\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mactive_vision.core\u001b[0m:\u001b[36msample_uncertain\u001b[0m:\u001b[36m119\u001b[0m - \u001b[1mGetting top 30 low confidence samples\u001b[0m\n"
     ]
    },
    {
     "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>filepath</th>\n",
       "      <th>pred_label</th>\n",
       "      <th>pred_conf</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1215</th>\n",
       "      <td>data/imagenette/0/01228.jpg</td>\n",
       "      <td>English springer</td>\n",
       "      <td>0.107802</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1970</th>\n",
       "      <td>data/imagenette/3/01991.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.108339</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7657</th>\n",
       "      <td>data/imagenette/6/07739.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.110674</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7257</th>\n",
       "      <td>data/imagenette/8/07335.jpg</td>\n",
       "      <td>parachute</td>\n",
       "      <td>0.110774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5385</th>\n",
       "      <td>data/imagenette/7/05443.jpg</td>\n",
       "      <td>tench</td>\n",
       "      <td>0.111138</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4898</th>\n",
       "      <td>data/imagenette/7/04950.jpg</td>\n",
       "      <td>garbage truck</td>\n",
       "      <td>0.111410</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7857</th>\n",
       "      <td>data/imagenette/6/07944.jpg</td>\n",
       "      <td>church</td>\n",
       "      <td>0.112245</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9194</th>\n",
       "      <td>data/imagenette/5/09291.jpg</td>\n",
       "      <td>garbage truck</td>\n",
       "      <td>0.112378</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7048</th>\n",
       "      <td>data/imagenette/8/07125.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.112463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5366</th>\n",
       "      <td>data/imagenette/7/05424.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.112498</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1477</th>\n",
       "      <td>data/imagenette/0/01492.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.112623</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2672</th>\n",
       "      <td>data/imagenette/3/02701.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.113079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2369</th>\n",
       "      <td>data/imagenette/3/02392.jpg</td>\n",
       "      <td>church</td>\n",
       "      <td>0.113294</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1882</th>\n",
       "      <td>data/imagenette/0/01902.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.113302</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2647</th>\n",
       "      <td>data/imagenette/3/02676.jpg</td>\n",
       "      <td>church</td>\n",
       "      <td>0.113753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4623</th>\n",
       "      <td>data/imagenette/9/04673.jpg</td>\n",
       "      <td>golf ball</td>\n",
       "      <td>0.113945</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>158</th>\n",
       "      <td>data/imagenette/2/00161.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.114293</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4711</th>\n",
       "      <td>data/imagenette/7/04762.jpg</td>\n",
       "      <td>church</td>\n",
       "      <td>0.114443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5409</th>\n",
       "      <td>data/imagenette/7/05468.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.114583</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1469</th>\n",
       "      <td>data/imagenette/0/01484.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.114606</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2042</th>\n",
       "      <td>data/imagenette/3/02063.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.114819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5182</th>\n",
       "      <td>data/imagenette/7/05237.jpg</td>\n",
       "      <td>gas pump</td>\n",
       "      <td>0.115313</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2184</th>\n",
       "      <td>data/imagenette/3/02207.jpg</td>\n",
       "      <td>church</td>\n",
       "      <td>0.115595</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5062</th>\n",
       "      <td>data/imagenette/7/05114.jpg</td>\n",
       "      <td>gas pump</td>\n",
       "      <td>0.115671</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1822</th>\n",
       "      <td>data/imagenette/0/01842.jpg</td>\n",
       "      <td>gas pump</td>\n",
       "      <td>0.115793</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2133</th>\n",
       "      <td>data/imagenette/3/02155.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.115859</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5019</th>\n",
       "      <td>data/imagenette/7/05071.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.116108</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5401</th>\n",
       "      <td>data/imagenette/7/05459.jpg</td>\n",
       "      <td>French horn</td>\n",
       "      <td>0.116387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2483</th>\n",
       "      <td>data/imagenette/3/02507.jpg</td>\n",
       "      <td>chain saw</td>\n",
       "      <td>0.116476</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2023</th>\n",
       "      <td>data/imagenette/3/02044.jpg</td>\n",
       "      <td>cassette player</td>\n",
       "      <td>0.116520</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         filepath        pred_label  pred_conf\n",
       "1215  data/imagenette/0/01228.jpg  English springer   0.107802\n",
       "1970  data/imagenette/3/01991.jpg       French horn   0.108339\n",
       "7657  data/imagenette/6/07739.jpg         chain saw   0.110674\n",
       "7257  data/imagenette/8/07335.jpg         parachute   0.110774\n",
       "5385  data/imagenette/7/05443.jpg             tench   0.111138\n",
       "4898  data/imagenette/7/04950.jpg     garbage truck   0.111410\n",
       "7857  data/imagenette/6/07944.jpg            church   0.112245\n",
       "9194  data/imagenette/5/09291.jpg     garbage truck   0.112378\n",
       "7048  data/imagenette/8/07125.jpg       French horn   0.112463\n",
       "5366  data/imagenette/7/05424.jpg   cassette player   0.112498\n",
       "1477  data/imagenette/0/01492.jpg         chain saw   0.112623\n",
       "2672  data/imagenette/3/02701.jpg         chain saw   0.113079\n",
       "2369  data/imagenette/3/02392.jpg            church   0.113294\n",
       "1882  data/imagenette/0/01902.jpg       French horn   0.113302\n",
       "2647  data/imagenette/3/02676.jpg            church   0.113753\n",
       "4623  data/imagenette/9/04673.jpg         golf ball   0.113945\n",
       "158   data/imagenette/2/00161.jpg   cassette player   0.114293\n",
       "4711  data/imagenette/7/04762.jpg            church   0.114443\n",
       "5409  data/imagenette/7/05468.jpg   cassette player   0.114583\n",
       "1469  data/imagenette/0/01484.jpg         chain saw   0.114606\n",
       "2042  data/imagenette/3/02063.jpg         chain saw   0.114819\n",
       "5182  data/imagenette/7/05237.jpg          gas pump   0.115313\n",
       "2184  data/imagenette/3/02207.jpg            church   0.115595\n",
       "5062  data/imagenette/7/05114.jpg          gas pump   0.115671\n",
       "1822  data/imagenette/0/01842.jpg          gas pump   0.115793\n",
       "2133  data/imagenette/3/02155.jpg         chain saw   0.115859\n",
       "5019  data/imagenette/7/05071.jpg         chain saw   0.116108\n",
       "5401  data/imagenette/7/05459.jpg       French horn   0.116387\n",
       "2483  data/imagenette/3/02507.jpg         chain saw   0.116476\n",
       "2023  data/imagenette/3/02044.jpg   cassette player   0.116520"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "uncertain_df = al.sample_uncertain(pred_df, num_samples=30)\n",
    "uncertain_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Label the uncertain samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "al.label(uncertain_df, output_filename=\"uncertain\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labeled_df = pd.read_parquet(\"uncertain.parquet\")\n",
    "\n",
    "labeled_df\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "al.add_to_train_set(labeled_df, output_filename=\"active_labeled\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.12.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}