first_submission.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from pathlib import Path\n",
    "\n",
    "from fastai import *\n",
    "from fastai.vision import *\n",
    "\n",
    "import json\n",
    "\n",
    "from utils import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "NUM_SAMPLES_PER_CLASS = 1_000\n",
    "NUM_VAL = 50 * 340"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "PATH = Path('data/txt')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generate data from csvs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "PATH.mkdir(exist_ok=True)\n",
    "(PATH/'train').mkdir(exist_ok=True)\n",
    "(PATH/'test').mkdir(exist_ok=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_train_txts_from_df(path):\n",
    "    df = pd.read_csv(path)\n",
    "    klass = '_'.join(path.stem.split())\n",
    "    (PATH/'train'/klass).mkdir(exist_ok=True)\n",
    "    for row in df.sample(NUM_SAMPLES_PER_CLASS).iterrows():\n",
    "        example = {\n",
    "            'countrycode': row[1].countrycode,\n",
    "            'drawing': json.loads(row[1].drawing),\n",
    "            'key_id': row[1].key_id,\n",
    "            'recognized': row[1].recognized\n",
    "        }\n",
    "        with open(PATH/'train'/klass/f'{example[\"key_id\"]}.txt', mode='w') as f: json.dump(example, f)\n",
    "\n",
    "def create_test_txts_from_df(path):\n",
    "    df = pd.read_csv(path)\n",
    "    for row in df.iterrows():\n",
    "        example = {\n",
    "            'countrycode': row[1].countrycode,\n",
    "            'drawing': json.loads(row[1].drawing),\n",
    "            'key_id': row[1].key_id\n",
    "        }\n",
    "        with open(PATH/'test'/f'{example[\"key_id\"]}.txt', mode='w') as f: json.dump(example, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 29.2 s, sys: 2.38 s, total: 31.6 s\n",
      "Wall time: 31.7 s\n"
     ]
    }
   ],
   "source": [
    "%time create_test_txts_from_df('data/test_simplified.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 5min 29s, sys: 18.6 s, total: 5min 48s\n",
      "Wall time: 5min 44s\n"
     ]
    }
   ],
   "source": [
    "%time for p in Path('data/train').iterdir(): create_train_txts_from_df(p)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Create ImageDataBunch using the data block API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "sz = 128\n",
    "bs = 640"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_func(path):\n",
    "    with open(path) as f: j = json.load(f)\n",
    "    drawing = list2drawing(j['drawing'], size=sz)\n",
    "    tensor = drawing2tensor(drawing)\n",
    "    return Image(tensor.div_(255))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "item_list = ItemList.from_folder(PATH/'train', create_func=create_func)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "idxs = np.arange(item_list.items.shape[0])\n",
    "np.random.shuffle(idxs)\n",
    "val_idxs = idxs[:NUM_VAL]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "item_lists = item_list.split_by_idx(val_idxs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# label_lists = item_lists.label_from_folder()\n",
    "# pd.to_pickle(label_lists.train.y.classes, 'data/classes.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "classes = pd.read_pickle('data/classes.pkl')\n",
    "\n",
    "label_lists = item_lists.label_from_folder(classes=classes)\n",
    "\n",
    "test_items = ItemList.from_folder(PATH/'test', create_func=create_func)\n",
    "label_lists.add_test(test_items);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dl = DataLoader(label_lists.train, bs, True, num_workers=12)\n",
    "valid_dl = DataLoader(label_lists.valid, 2*bs, False, num_workers=12)\n",
    "test_dl = DataLoader(label_lists.test, 2*bs, False, num_workers=12)\n",
    "\n",
    "data_bunch = ImageDataBunch(train_dl, valid_dl, test_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pd.to_pickle(data_bunch.batch_stats(), f'data/batch_stats_{sz}.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_stats = pd.read_pickle(f'data/batch_stats_{sz}.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_bunch.normalize(batch_stats);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x720 with 16 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data_bunch.show_batch(rows=4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "name = f'res34-{sz}'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = create_cnn(data_bunch, models.resnet34, metrics=[accuracy, map3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total time: 06:00\n",
      "epoch  train_loss  valid_loss  accuracy  map3    \n",
      "1      1.937478    1.607339    0.608647  0.695402  (03:01)\n",
      "2      1.658560    1.443055    0.652294  0.731284  (02:59)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn.fit_one_cycle(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save(f'{name}-stage-1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total time: 24:25\n",
      "epoch  train_loss  valid_loss  accuracy  map3    \n",
      "1      1.448672    1.383536    0.658706  0.739206  (04:04)\n",
      "2      1.358841    1.285606    0.681059  0.757471  (04:04)\n",
      "3      1.165717    1.180751    0.707647  0.780500  (04:04)\n",
      "4      0.927250    1.081052    0.732294  0.800314  (04:03)\n",
      "5      0.639683    1.044819    0.744294  0.810804  (04:03)\n",
      "6      0.454999    1.067124    0.744647  0.810461  (04:03)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn.fit_one_cycle(6, max_lr=6e-4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save(f'{name}-stage-2')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load(f'{name}-stage-2');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds, _ = learn.get_preds(ds_type=DatasetType.Test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "create_submission(preds, data_bunch.test_dl, name, classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "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>key_id</th>\n",
       "      <th>word</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>9221337500464606</td>\n",
       "      <td>sock shoe foot</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>9950013608613660</td>\n",
       "      <td>leaf feather stitches</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>9005540600650567</td>\n",
       "      <td>broom paintbrush rake</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>9031090525591545</td>\n",
       "      <td>chair giraffe toilet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9896706467105420</td>\n",
       "      <td>flamingo swan bird</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             key_id                   word\n",
       "0  9221337500464606         sock shoe foot\n",
       "1  9950013608613660  leaf feather stitches\n",
       "2  9005540600650567  broom paintbrush rake\n",
       "3  9031090525591545   chair giraffe toilet\n",
       "4  9896706467105420     flamingo swan bird"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv(f'subs/{name}.csv.gz').head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████| 1.52M/1.52M [00:17<00:00, 93.0kB/s]\n",
      "Successfully submitted to Quick, Draw! Doodle Recognition Challenge"
     ]
    }
   ],
   "source": [
    "# !kaggle competitions submit -c quickdraw-doodle-recognition -f subs/{name}.csv.gz -m \"{name}\""
   ]
  }
 ],
 "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}