code/Tosho_Book_Representation_Tensorboard.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualize Book Representation in TensborBoard"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import gensim\n",
    "import torch\n",
    "import torchvision\n",
    "from tensorboardX import SummaryWriter\n",
    "from PIL import Image\n",
    "from pathlib import Path\n",
    "import shutil"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert ISBN to Book ID"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(891, 3)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "desc_pd = pd.read_csv('data/tosho_processed_clean.csv.bz2', sep='\\t', compression='bz2')\n",
    "desc_pd.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1187, 4)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_df = pd.read_csv('./data/tosho_raw.csv.bz2', sep='\\t', compression='bz2')\n",
    "id_df.drop_duplicates(inplace=True)\n",
    "id_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(891, 6)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_df = pd.merge(desc_pd, id_df, on=['id'])\n",
    "id_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def modulus11weight10to2(c9):\n",
    "    sum_ = 0\n",
    "    for i in range(len(c9)):\n",
    "        try:\n",
    "            c = int(c9[i])\n",
    "        except ValueError:\n",
    "            return False\n",
    "        sum_ += (10 - i) * c\n",
    "\n",
    "    result = 11 - (sum_ % 11)\n",
    "    \n",
    "    if result == 11:\n",
    "        return '0'\n",
    "    elif result == 10:\n",
    "        return 'X'\n",
    "    else:\n",
    "        return str(result)\n",
    "    \n",
    "def encode13to10(isbn13):\n",
    "    prefix = isbn13[3:-1]\n",
    "    check_digit = modulus11weight10to2(prefix)\n",
    "    isbn10 = prefix + check_digit\n",
    "    return isbn10\n",
    "\n",
    "def convert_isbn(isbn):\n",
    "    if len(isbn) == 13:\n",
    "        return encode13to10(isbn)\n",
    "    else:\n",
    "        return isbn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "id_df['isbn'] = id_df['isbn_raw'].apply(lambda x: convert_isbn(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "book_cover_list = !ls data/book_img_original"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(870, 1)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "book_cover_pd = pd.DataFrame(book_cover_list, columns=['file_name'])\n",
    "book_cover_pd.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "book_cover_pd['isbn'] = book_cover_pd.apply(lambda x: x.file_name.replace('.jpg', ''), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(891, 3)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_isbn_mapping_df = pd.merge(id_df[['id', 'isbn']], book_cover_pd[['isbn', 'file_name']], on=['isbn'])\n",
    "id_isbn_mapping_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>id</th>\n",
       "      <th>isbn</th>\n",
       "      <th>file_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3901637</td>\n",
       "      <td>4577028603</td>\n",
       "      <td>4577028603.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3901633</td>\n",
       "      <td>4772101896</td>\n",
       "      <td>4772101896.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3901632</td>\n",
       "      <td>4772101829</td>\n",
       "      <td>4772101829.jpg</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        id        isbn       file_name\n",
       "0  3901637  4577028603  4577028603.jpg\n",
       "1  3901633  4772101896  4772101896.jpg\n",
       "2  3901632  4772101829  4772101829.jpg"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_isbn_mapping_df.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Rename image filenames and copy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "dst_directory = 'data/book_img'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "!mkdir -p {dst_directory}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def move_book_covers(filename, book_id):\n",
    "    dst = 'data/book_img'\n",
    "    try:\n",
    "        shutil.copy('data/book_img_original/'+filename, dst_directory+'/'+str(book_id)+'.jpg')\n",
    "        return 1\n",
    "    except:\n",
    "        return 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "id_isbn_mapping_df['result'] = id_isbn_mapping_df.apply(lambda x: move_book_covers(x.file_name, x.id), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    891\n",
       "Name: result, dtype: int64"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_isbn_mapping_df['result'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load book representation model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/smart_open/smart_open_lib.py:398: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n",
      "  'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n",
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:4: DeprecationWarning: Call to deprecated `wv` (Attribute will be removed in 4.0.0, use self instead).\n",
      "  after removing the cwd from sys.path.\n",
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:5: DeprecationWarning: Call to deprecated `wv` (Attribute will be removed in 4.0.0, use self instead).\n",
      "  \"\"\"\n"
     ]
    }
   ],
   "source": [
    "writer = SummaryWriter()\n",
    "model = gensim.models.keyedvectors.KeyedVectors.load_word2vec_format('model/book2vec', binary=False)\n",
    "\n",
    "weights = model.wv.vectors\n",
    "labels = model.wv.index2word"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert Book ID's to Book Titles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels_title = [desc_pd[desc_pd.id == int(i)]['title'].values[0] for i in labels]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create Book Cover Sprites"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_size = 128\n",
    "\n",
    "transform = torchvision.transforms.Compose([\n",
    "    torchvision.transforms.Resize((img_size, img_size)),\n",
    "    torchvision.transforms.ToTensor(),\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "images = torch.zeros(0)\n",
    "\n",
    "for i, img in enumerate(Path(dst_directory).glob(\"*.jpg\")):\n",
    "    img_v = transform(Image.open(img))\n",
    "    images = torch.cat((images, img_v))\n",
    "\n",
    "images = images.view(len(labels), 3, img_size, img_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Write to Tensorboard format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "writer.add_embedding(torch.FloatTensor(weights), label_img=images, metadata=labels_title)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. Run Tensorboard  \n",
    "```bash\n",
    "tensorboard --logdir=runs\n",
    "```\n",
    "2. Go to `localhost:6006` on your web browser"
   ]
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}