dataset-exploration.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "1"
    }
   },
   "source": [
    "# Match Cutting\n",
    "### Dataset exploration\n",
    "We'll use some of the classes in the repo to read and analyze data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "nterop": {
     "id": "2"
    }
   },
   "outputs": [],
   "source": [
    "import json\n",
    "from typing import Set\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "from matchcut.data import Dataset"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "3"
    }
   },
   "source": [
    "### Title set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {
    "nterop": {
     "id": "4"
    }
   },
   "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>IMDB ID</th>\n",
       "      <th>title</th>\n",
       "      <th>genres</th>\n",
       "      <th>country</th>\n",
       "      <th>split</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tt0050706</td>\n",
       "      <td>Mon Oncle (1958)</td>\n",
       "      <td>Comedy</td>\n",
       "      <td>France</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>tt0059592</td>\n",
       "      <td>Pierrot le Fou (1965)</td>\n",
       "      <td>Crime, Drama, Romance</td>\n",
       "      <td>France</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tt0061722</td>\n",
       "      <td>The Graduate (1967)</td>\n",
       "      <td>Comedy, Drama, Romance</td>\n",
       "      <td>USA</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>tt0061781</td>\n",
       "      <td>The Firemen's Ball (1967)</td>\n",
       "      <td>Comedy, Drama</td>\n",
       "      <td>Czechoslovakia</td>\n",
       "      <td>validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tt0066921</td>\n",
       "      <td>A Clockwork Orange (1971)</td>\n",
       "      <td>Crime, Drama, Sci-Fi</td>\n",
       "      <td>UK</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>95</th>\n",
       "      <td>tt5827496</td>\n",
       "      <td>At Cafe 6 (2016)</td>\n",
       "      <td>Comedy, Romance</td>\n",
       "      <td>Taiwan</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96</th>\n",
       "      <td>tt5866930</td>\n",
       "      <td>The Adventurers (2017)</td>\n",
       "      <td>Action, Adventure, Crime, Drama</td>\n",
       "      <td>China</td>\n",
       "      <td>validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>97</th>\n",
       "      <td>tt6157626</td>\n",
       "      <td>Legend of the Demon Cat (2017)</td>\n",
       "      <td>Drama, Fantasy, Horror, Mystery, Romance</td>\n",
       "      <td>China</td>\n",
       "      <td>validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98</th>\n",
       "      <td>tt6298600</td>\n",
       "      <td>The Miracles of the Namiya General Store (2017)</td>\n",
       "      <td>Drama</td>\n",
       "      <td>Japan</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>Bad Genius (2017)</td>\n",
       "      <td>Comedy, Crime, Drama, Thriller</td>\n",
       "      <td>Thailand</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>100 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      IMDB ID                                            title  \\\n",
       "0   tt0050706                                 Mon Oncle (1958)   \n",
       "1   tt0059592                            Pierrot le Fou (1965)   \n",
       "2   tt0061722                              The Graduate (1967)   \n",
       "3   tt0061781                        The Firemen's Ball (1967)   \n",
       "4   tt0066921                        A Clockwork Orange (1971)   \n",
       "..        ...                                              ...   \n",
       "95  tt5827496                                 At Cafe 6 (2016)   \n",
       "96  tt5866930                           The Adventurers (2017)   \n",
       "97  tt6157626                   Legend of the Demon Cat (2017)   \n",
       "98  tt6298600  The Miracles of the Namiya General Store (2017)   \n",
       "99  tt6788942                                Bad Genius (2017)   \n",
       "\n",
       "                                      genres         country       split  \n",
       "0                                     Comedy          France       train  \n",
       "1                      Crime, Drama, Romance          France       train  \n",
       "2                     Comedy, Drama, Romance             USA       train  \n",
       "3                              Comedy, Drama  Czechoslovakia  validation  \n",
       "4                       Crime, Drama, Sci-Fi              UK       train  \n",
       "..                                       ...             ...         ...  \n",
       "95                           Comedy, Romance          Taiwan        test  \n",
       "96           Action, Adventure, Crime, Drama           China  validation  \n",
       "97  Drama, Fantasy, Horror, Mystery, Romance           China  validation  \n",
       "98                                     Drama           Japan       train  \n",
       "99            Comedy, Crime, Drama, Thriller        Thailand       train  \n",
       "\n",
       "[100 rows x 5 columns]"
      ]
     },
     "execution_count": 122,
     "metadata": {
      "nterop": {
       "id": "145"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_titles = pd.read_csv('data/imdb-title-set.csv')\n",
    "df_titles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "nterop": {
     "id": "79"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "nterop": {
       "id": "141"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_titles.genres.apply(lambda x: x.split(\", \")).explode().value_counts()[::-1].plot(kind='barh')\n",
    "_ = plt.xlabel(\"Number of titles\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "nterop": {
     "id": "81"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "nterop": {
       "id": "142"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "_ = df_titles.country.value_counts()[::-1].plot(kind='barh')\n",
    "_ = plt.xlabel(\"Number of titles\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "nterop": {
     "id": "83"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count     100.000000\n",
       "mean     2002.800000\n",
       "std        13.944334\n",
       "min      1958.000000\n",
       "25%      1996.500000\n",
       "50%      2008.000000\n",
       "75%      2013.000000\n",
       "max      2017.000000\n",
       "Name: 0, dtype: float64"
      ]
     },
     "execution_count": 71,
     "metadata": {
      "nterop": {
       "id": "143"
      }
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "nterop": {
       "id": "144"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "release_years = (\n",
    "    df_titles\n",
    "    .title\n",
    "    .str.extract(\"\\((\\d+)\\)\")[0]\n",
    "    .astype(int)\n",
    ")\n",
    "release_years.hist()\n",
    "_ = plt.xlabel('Year')\n",
    "_ = plt.ylabel(\"Number of titles\")\n",
    "release_years.describe()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "6"
    }
   },
   "source": [
    "### get all paired data for a task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "nterop": {
     "id": "7"
    }
   },
   "outputs": [],
   "source": [
    "def get_df(task: str, source: str = None) -> pd.DataFrame:\n",
    "    \"\"\"\n",
    "    Read a dataframe where `task` is one of frame or motion.\n",
    "    if `source` is not provided all the data is returned,\n",
    "    if it is then the data is filtered down to the pairs\n",
    "    retrieved by one of the 4 heuristics.\n",
    "    \"\"\"\n",
    "    ds = Dataset(\n",
    "        task=task,\n",
    "        split=\"train\",  # ignored\n",
    "        encoder_name=None,\n",
    "        agg_name=None,\n",
    "        source=source,\n",
    "    )\n",
    "    return pd.DataFrame(p.__dict__ for p in ds.pairs_labeled_all)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "8"
    }
   },
   "source": [
    "#### convert to a dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "nterop": {
     "id": "9"
    }
   },
   "outputs": [],
   "source": [
    "df_frame = get_df('frame')\n",
    "df_motion = get_df('motion')\n",
    "df_h1 = get_df('frame', 'heuristic 1: faces')\n",
    "df_h2 = get_df('frame', 'heuristic 2: instance segmentation')\n",
    "df_h3 = get_df('motion', 'heuristic 3: Farneback')\n",
    "df_h4 = get_df('motion', 'heuristic 4: RAFT')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "nterop": {
     "id": "10"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9985"
      ]
     },
     "execution_count": 57,
     "metadata": {
      "nterop": {
       "id": "134"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(df_frame)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "nterop": {
     "id": "12"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9320"
      ]
     },
     "execution_count": 58,
     "metadata": {
      "nterop": {
       "id": "135"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(df_motion)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "nterop": {
     "id": "136"
    }
   },
   "outputs": [],
   "source": [
    "len(get_unique_shots(df_frame))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "nterop": {
     "id": "137"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "12993"
      ]
     },
     "execution_count": 62,
     "metadata": {
      "nterop": {
       "id": "138"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def get_unique_shots(df):\n",
    "    return set(x for _, r in df.iterrows() for x in ((r.imdb_id, r.shot1_idx), (r.imdb_id, r.shot2_idx)))\n",
    "\n",
    "len(get_unique_shots(df_frame) | get_unique_shots(df_motion))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "nterop": {
     "id": "14"
    }
   },
   "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>imdb_id</th>\n",
       "      <th>shot1_idx</th>\n",
       "      <th>shot2_idx</th>\n",
       "      <th>label</th>\n",
       "      <th>source_info</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1789</th>\n",
       "      <td>tt2267998</td>\n",
       "      <td>582</td>\n",
       "      <td>1738</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1817</th>\n",
       "      <td>tt2267998</td>\n",
       "      <td>935</td>\n",
       "      <td>1146</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2285</th>\n",
       "      <td>tt0178868</td>\n",
       "      <td>195</td>\n",
       "      <td>349</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017</th>\n",
       "      <td>tt3672840</td>\n",
       "      <td>1160</td>\n",
       "      <td>1565</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6153</th>\n",
       "      <td>tt0427954</td>\n",
       "      <td>238</td>\n",
       "      <td>1856</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        imdb_id  shot1_idx  shot2_idx  label  \\\n",
       "1789  tt2267998        582       1738  False   \n",
       "1817  tt2267998        935       1146  False   \n",
       "2285  tt0178868        195        349  False   \n",
       "2017  tt3672840       1160       1565  False   \n",
       "6153  tt0427954        238       1856  False   \n",
       "\n",
       "                                            source_info  \n",
       "1789  SourceInfo(sources_and_scores=(('heuristic 3: ...  \n",
       "1817  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "2285  SourceInfo(sources_and_scores=(('heuristic 3: ...  \n",
       "2017  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "6153  SourceInfo(sources_and_scores=(('heuristic 4: ...  "
      ]
     },
     "execution_count": 72,
     "metadata": {
      "nterop": {
       "id": "88"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_motion.sample(5, random_state=0)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "89"
    }
   },
   "source": [
    "### Dataset statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "nterop": {
     "id": "90"
    }
   },
   "outputs": [],
   "source": [
    "def get_stats(df: pd.DataFrame, task: str):\n",
    "    pairs = json.load(open(f\"data/dataset-{task}.json\"))\n",
    "    pa_cnt = pd.Series(\n",
    "        sum(pair[\"annotator_labels\"].values()) in (0, 3)\n",
    "        for pair in pairs\n",
    "    ).sum()\n",
    "    return {\n",
    "        \"Task\": task,\n",
    "        \"Annotated pairs\": len(df),\n",
    "        \"Positive pairs (majority label)\": df.label.sum(),\n",
    "        \"Positive rate\": df.label.mean(),\n",
    "        \"Perfect agreement pairs\": pa_cnt,\n",
    "        \"Perfect agreement\": pa_cnt / len(df),\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "nterop": {
     "id": "91"
    }
   },
   "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>Task</th>\n",
       "      <th>frame</th>\n",
       "      <th>motion</th>\n",
       "      <th>overall</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Annotated pairs</th>\n",
       "      <td>9985.000</td>\n",
       "      <td>9320.000</td>\n",
       "      <td>19305.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Positive pairs (majority label)</th>\n",
       "      <td>867.000</td>\n",
       "      <td>927.000</td>\n",
       "      <td>1794.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Positive rate</th>\n",
       "      <td>0.087</td>\n",
       "      <td>0.099</td>\n",
       "      <td>0.092929</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Perfect agreement pairs</th>\n",
       "      <td>8373.000</td>\n",
       "      <td>7027.000</td>\n",
       "      <td>15400.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Perfect agreement</th>\n",
       "      <td>0.839</td>\n",
       "      <td>0.754</td>\n",
       "      <td>0.797721</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Task                                frame    motion       overall\n",
       "Annotated pairs                  9985.000  9320.000  19305.000000\n",
       "Positive pairs (majority label)   867.000   927.000   1794.000000\n",
       "Positive rate                       0.087     0.099      0.092929\n",
       "Perfect agreement pairs          8373.000  7027.000  15400.000000\n",
       "Perfect agreement                   0.839     0.754      0.797721"
      ]
     },
     "execution_count": 82,
     "metadata": {
      "nterop": {
       "id": "146"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_stats = pd.DataFrame([\n",
    "    get_stats(df_frame, \"frame\"),\n",
    "    get_stats(df_motion, \"motion\"),\n",
    "])\n",
    "df_stats = df_stats.set_index(\"Task\").transpose().round(3)\n",
    "df_stats = df_stats.assign(overall=df_stats.frame + df_stats.motion)\n",
    "df_stats.loc['Positive rate', 'overall'] = df_stats.loc['Positive pairs (majority label)', 'overall'] / df_stats.loc['Annotated pairs', 'overall']\n",
    "df_stats.loc['Perfect agreement', 'overall'] = df_stats.loc['Perfect agreement pairs', 'overall'] / df_stats.loc['Annotated pairs', 'overall']\n",
    "df_stats"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "16"
    }
   },
   "source": [
    "### heuristics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "nterop": {
     "id": "17"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.08683024536805208"
      ]
     },
     "execution_count": 83,
     "metadata": {
      "nterop": {
       "id": "147"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_frame.label.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "nterop": {
     "id": "19"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.09946351931330472"
      ]
     },
     "execution_count": 84,
     "metadata": {
      "nterop": {
       "id": "148"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_motion.label.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "nterop": {
     "id": "21"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0118"
      ]
     },
     "execution_count": 85,
     "metadata": {
      "nterop": {
       "id": "149"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_h1.label.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "nterop": {
     "id": "23"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.1616"
      ]
     },
     "execution_count": 86,
     "metadata": {
      "nterop": {
       "id": "150"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_h2.label.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "nterop": {
     "id": "25"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.1086"
      ]
     },
     "execution_count": 87,
     "metadata": {
      "nterop": {
       "id": "151"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_h3.label.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "nterop": {
     "id": "27"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0988"
      ]
     },
     "execution_count": 88,
     "metadata": {
      "nterop": {
       "id": "152"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_h4.label.mean()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "29"
    }
   },
   "source": [
    "### title-level positive rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "nterop": {
     "id": "30"
    }
   },
   "outputs": [],
   "source": [
    "def title_level_positive_rate(df: pd.DataFrame) -> pd.Series:\n",
    "    return df.groupby('imdb_id').label.mean().describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "nterop": {
     "id": "31"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100.000000\n",
       "mean       0.086842\n",
       "std        0.060670\n",
       "min        0.000000\n",
       "25%        0.040000\n",
       "50%        0.070707\n",
       "75%        0.120000\n",
       "max        0.320000\n",
       "Name: label, dtype: float64"
      ]
     },
     "execution_count": 90,
     "metadata": {
      "nterop": {
       "id": "153"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title_level_positive_rate(df_frame)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "nterop": {
     "id": "33"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100.000000\n",
       "mean       0.100124\n",
       "std        0.074155\n",
       "min        0.000000\n",
       "25%        0.043276\n",
       "50%        0.081220\n",
       "75%        0.147883\n",
       "max        0.340659\n",
       "Name: label, dtype: float64"
      ]
     },
     "execution_count": 91,
     "metadata": {
      "nterop": {
       "id": "154"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title_level_positive_rate(df_motion)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "35"
    }
   },
   "source": [
    "#### by heuristic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "nterop": {
     "id": "36"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100.000000\n",
       "mean       0.011800\n",
       "std        0.023285\n",
       "min        0.000000\n",
       "25%        0.000000\n",
       "50%        0.000000\n",
       "75%        0.020000\n",
       "max        0.120000\n",
       "Name: label, dtype: float64"
      ]
     },
     "execution_count": 92,
     "metadata": {
      "nterop": {
       "id": "155"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title_level_positive_rate(df_h1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "nterop": {
     "id": "38"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100.000000\n",
       "mean       0.161600\n",
       "std        0.113944\n",
       "min        0.000000\n",
       "25%        0.080000\n",
       "50%        0.140000\n",
       "75%        0.220000\n",
       "max        0.580000\n",
       "Name: label, dtype: float64"
      ]
     },
     "execution_count": 93,
     "metadata": {
      "nterop": {
       "id": "156"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title_level_positive_rate(df_h2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "nterop": {
     "id": "40"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100.0000\n",
       "mean       0.1086\n",
       "std        0.0860\n",
       "min        0.0000\n",
       "25%        0.0400\n",
       "50%        0.1000\n",
       "75%        0.1400\n",
       "max        0.3800\n",
       "Name: label, dtype: float64"
      ]
     },
     "execution_count": 94,
     "metadata": {
      "nterop": {
       "id": "157"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title_level_positive_rate(df_h3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "nterop": {
     "id": "42"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100.000000\n",
       "mean       0.098800\n",
       "std        0.088435\n",
       "min        0.000000\n",
       "25%        0.040000\n",
       "50%        0.080000\n",
       "75%        0.140000\n",
       "max        0.380000\n",
       "Name: label, dtype: float64"
      ]
     },
     "execution_count": 95,
     "metadata": {
      "nterop": {
       "id": "158"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title_level_positive_rate(df_h4)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "44"
    }
   },
   "source": [
    "### shot time ranges"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "nterop": {
     "id": "45"
    }
   },
   "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>imdb_id</th>\n",
       "      <th>shot_idx</th>\n",
       "      <th>start</th>\n",
       "      <th>end</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tt0050706</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>27.583333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>tt0050706</td>\n",
       "      <td>6</td>\n",
       "      <td>123.541667</td>\n",
       "      <td>137.916667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tt0050706</td>\n",
       "      <td>7</td>\n",
       "      <td>137.916667</td>\n",
       "      <td>152.958333</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     imdb_id  shot_idx       start         end\n",
       "0  tt0050706         0    0.000000   27.583333\n",
       "1  tt0050706         6  123.541667  137.916667\n",
       "2  tt0050706         7  137.916667  152.958333"
      ]
     },
     "execution_count": 96,
     "metadata": {
      "nterop": {
       "id": "159"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_time_ranges = pd.read_csv('data/shot-time-ranges.csv')\n",
    "df_time_ranges.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "nterop": {
     "id": "106"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "nterop": {
       "id": "160"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_time_ranges.groupby('imdb_id').size().hist()\n",
    "_ = plt.xlabel('Number of unique shots')\n",
    "_ = plt.ylabel('Number of titles')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "109"
    }
   },
   "source": [
    "##### shot duration stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "nterop": {
     "id": "110"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    21205.000000\n",
       "mean         8.173633\n",
       "std         15.135620\n",
       "min          0.240000\n",
       "25%          2.083333\n",
       "50%          3.878875\n",
       "75%          8.091416\n",
       "max        384.500000\n",
       "Name: duration, dtype: float64"
      ]
     },
     "execution_count": 99,
     "metadata": {
      "nterop": {
       "id": "161"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_time_ranges_duration = df_time_ranges.assign(duration=df_time_ranges.end - df_time_ranges.start)\n",
    "df_time_ranges_duration = df_time_ranges_duration.merge(df_titles.rename(columns={\"IMDB ID\": \"imdb_id\"}))\n",
    "df_time_ranges_duration.duration.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "nterop": {
     "id": "112"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "nterop": {
       "id": "162"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "(\n",
    "    df_time_ranges_duration\n",
    "    .assign(genre=df_time_ranges_duration.genres.str.split(\", \")).explode(\"genre\")\n",
    "    .boxplot(column=\"duration\", by=\"genre\", vert=False, figsize=(12, 8))\n",
    ")\n",
    "plt.title('Shot duration distribution by genre')\n",
    "plt.suptitle('')\n",
    "_ = plt.xlabel('Shot duration in seconds')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "47"
    }
   },
   "source": [
    "#### merge to get time ranges for pairs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "nterop": {
     "id": "48"
    }
   },
   "outputs": [],
   "source": [
    "def merge_time_ranges(df: pd.DataFrame, nbr: int) -> pd.DataFrame:\n",
    "    \"\"\"\n",
    "    `df` to append time ranges (start and end time).\n",
    "    `nbr` is either 1 or 2.\n",
    "    \"\"\"\n",
    "    return df.merge(\n",
    "        df_time_ranges.add_prefix(f\"shot{nbr}_\"),\n",
    "        left_on=['imdb_id', f'shot{nbr}_idx'],\n",
    "        right_on=[f'shot{nbr}_imdb_id', f'shot{nbr}_shot_idx'],\n",
    "    ).drop(columns=[f'shot{nbr}_shot_idx', f'shot{nbr}_imdb_id'])\n",
    "\n",
    "\n",
    "def add_time_ranges(df: pd.DataFrame) -> pd.DataFrame:\n",
    "    return merge_time_ranges(merge_time_ranges(df, 1), 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "nterop": {
     "id": "49"
    }
   },
   "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>imdb_id</th>\n",
       "      <th>shot1_idx</th>\n",
       "      <th>shot2_idx</th>\n",
       "      <th>label</th>\n",
       "      <th>source_info</th>\n",
       "      <th>shot1_start</th>\n",
       "      <th>shot1_end</th>\n",
       "      <th>shot2_start</th>\n",
       "      <th>shot2_end</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>92</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>91.716623</td>\n",
       "      <td>103.019581</td>\n",
       "      <td>470.595112</td>\n",
       "      <td>476.225737</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>112</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>91.716623</td>\n",
       "      <td>103.019581</td>\n",
       "      <td>608.941650</td>\n",
       "      <td>613.988358</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>294</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>91.716623</td>\n",
       "      <td>103.019581</td>\n",
       "      <td>1502.959751</td>\n",
       "      <td>1506.546667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>500</td>\n",
       "      <td>True</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>91.716623</td>\n",
       "      <td>103.019581</td>\n",
       "      <td>2482.521641</td>\n",
       "      <td>2492.323099</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>530</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>91.716623</td>\n",
       "      <td>103.019581</td>\n",
       "      <td>2677.132719</td>\n",
       "      <td>2679.510094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9315</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1089</td>\n",
       "      <td>1193</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>4961.250000</td>\n",
       "      <td>4965.250000</td>\n",
       "      <td>5320.583333</td>\n",
       "      <td>5321.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9316</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1089</td>\n",
       "      <td>1380</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>4961.250000</td>\n",
       "      <td>4965.250000</td>\n",
       "      <td>5861.666667</td>\n",
       "      <td>5865.750000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9317</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1203</td>\n",
       "      <td>1380</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>5337.458333</td>\n",
       "      <td>5341.750000</td>\n",
       "      <td>5861.666667</td>\n",
       "      <td>5865.750000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9318</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1089</td>\n",
       "      <td>1426</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>4961.250000</td>\n",
       "      <td>4965.250000</td>\n",
       "      <td>6035.458333</td>\n",
       "      <td>6039.625000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9319</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1203</td>\n",
       "      <td>1426</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>5337.458333</td>\n",
       "      <td>5341.750000</td>\n",
       "      <td>6035.458333</td>\n",
       "      <td>6039.625000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>9320 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        imdb_id  shot1_idx  shot2_idx  label  \\\n",
       "0     tt0118799          6         92  False   \n",
       "1     tt0118799          6        112  False   \n",
       "2     tt0118799          6        294  False   \n",
       "3     tt0118799          6        500   True   \n",
       "4     tt0118799          6        530  False   \n",
       "...         ...        ...        ...    ...   \n",
       "9315  tt6788942       1089       1193  False   \n",
       "9316  tt6788942       1089       1380  False   \n",
       "9317  tt6788942       1203       1380  False   \n",
       "9318  tt6788942       1089       1426  False   \n",
       "9319  tt6788942       1203       1426  False   \n",
       "\n",
       "                                            source_info  shot1_start  \\\n",
       "0     SourceInfo(sources_and_scores=(('heuristic 3: ...    91.716623   \n",
       "1     SourceInfo(sources_and_scores=(('heuristic 3: ...    91.716623   \n",
       "2     SourceInfo(sources_and_scores=(('heuristic 3: ...    91.716623   \n",
       "3     SourceInfo(sources_and_scores=(('heuristic 3: ...    91.716623   \n",
       "4     SourceInfo(sources_and_scores=(('heuristic 3: ...    91.716623   \n",
       "...                                                 ...          ...   \n",
       "9315  SourceInfo(sources_and_scores=(('heuristic 4: ...  4961.250000   \n",
       "9316  SourceInfo(sources_and_scores=(('heuristic 4: ...  4961.250000   \n",
       "9317  SourceInfo(sources_and_scores=(('heuristic 4: ...  5337.458333   \n",
       "9318  SourceInfo(sources_and_scores=(('heuristic 4: ...  4961.250000   \n",
       "9319  SourceInfo(sources_and_scores=(('heuristic 4: ...  5337.458333   \n",
       "\n",
       "        shot1_end  shot2_start    shot2_end  \n",
       "0      103.019581   470.595112   476.225737  \n",
       "1      103.019581   608.941650   613.988358  \n",
       "2      103.019581  1502.959751  1506.546667  \n",
       "3      103.019581  2482.521641  2492.323099  \n",
       "4      103.019581  2677.132719  2679.510094  \n",
       "...           ...          ...          ...  \n",
       "9315  4965.250000  5320.583333  5321.333333  \n",
       "9316  4965.250000  5861.666667  5865.750000  \n",
       "9317  5341.750000  5861.666667  5865.750000  \n",
       "9318  4965.250000  6035.458333  6039.625000  \n",
       "9319  5341.750000  6035.458333  6039.625000  \n",
       "\n",
       "[9320 rows x 9 columns]"
      ]
     },
     "execution_count": 102,
     "metadata": {
      "nterop": {
       "id": "163"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "add_time_ranges(df_motion)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "51"
    }
   },
   "source": [
    "### splits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "nterop": {
     "id": "52"
    }
   },
   "outputs": [],
   "source": [
    "def get_split_ids(split: str) -> Set[str]:\n",
    "    return Dataset(split, \"frame\", None, None).imdb_ids"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "nterop": {
     "id": "53"
    }
   },
   "outputs": [],
   "source": [
    "splits = (\"train\", \"validation\", \"test\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {
    "nterop": {
     "id": "54"
    }
   },
   "outputs": [],
   "source": [
    "split_ids = {s: get_split_ids(s) for s in splits}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "nterop": {
     "id": "55"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['train', 'validation', 'test'])"
      ]
     },
     "execution_count": 106,
     "metadata": {
      "nterop": {
       "id": "164"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "split_ids.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "nterop": {
     "id": "57"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'tt0081505',\n",
       " 'tt0090257',\n",
       " 'tt0092603',\n",
       " 'tt0101700',\n",
       " 'tt0108656',\n",
       " 'tt0118799',\n",
       " 'tt0118845',\n",
       " 'tt0133093',\n",
       " 'tt0208092',\n",
       " 'tt0250494',\n",
       " 'tt0266697',\n",
       " 'tt0457430',\n",
       " 'tt0940709',\n",
       " 'tt0993846',\n",
       " 'tt1602620',\n",
       " 'tt2258281',\n",
       " 'tt3508840',\n",
       " 'tt3672840',\n",
       " 'tt5580036',\n",
       " 'tt5827496'}"
      ]
     },
     "execution_count": 107,
     "metadata": {
      "nterop": {
       "id": "165"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "split_ids['test']"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "59"
    }
   },
   "source": [
    "#### filter by split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {
    "nterop": {
     "id": "60"
    }
   },
   "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>imdb_id</th>\n",
       "      <th>shot1_idx</th>\n",
       "      <th>shot2_idx</th>\n",
       "      <th>label</th>\n",
       "      <th>source_info</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>265</th>\n",
       "      <td>tt1560747</td>\n",
       "      <td>2</td>\n",
       "      <td>139</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>266</th>\n",
       "      <td>tt1560747</td>\n",
       "      <td>2</td>\n",
       "      <td>166</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>267</th>\n",
       "      <td>tt1560747</td>\n",
       "      <td>2</td>\n",
       "      <td>402</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>268</th>\n",
       "      <td>tt1560747</td>\n",
       "      <td>49</td>\n",
       "      <td>54</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>269</th>\n",
       "      <td>tt1560747</td>\n",
       "      <td>54</td>\n",
       "      <td>61</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9315</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1203</td>\n",
       "      <td>1426</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9316</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1206</td>\n",
       "      <td>1367</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9317</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1505</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9318</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1558</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9319</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1721</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5594 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        imdb_id  shot1_idx  shot2_idx  label  \\\n",
       "265   tt1560747          2        139  False   \n",
       "266   tt1560747          2        166  False   \n",
       "267   tt1560747          2        402  False   \n",
       "268   tt1560747         49         54  False   \n",
       "269   tt1560747         54         61  False   \n",
       "...         ...        ...        ...    ...   \n",
       "9315  tt6788942       1203       1426  False   \n",
       "9316  tt6788942       1206       1367  False   \n",
       "9317  tt6788942       1367       1505  False   \n",
       "9318  tt6788942       1367       1558  False   \n",
       "9319  tt6788942       1367       1721  False   \n",
       "\n",
       "                                            source_info  \n",
       "265   SourceInfo(sources_and_scores=(('heuristic 3: ...  \n",
       "266   SourceInfo(sources_and_scores=(('heuristic 3: ...  \n",
       "267   SourceInfo(sources_and_scores=(('heuristic 3: ...  \n",
       "268   SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "269   SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "...                                                 ...  \n",
       "9315  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "9316  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "9317  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "9318  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "9319  SourceInfo(sources_and_scores=(('heuristic 4: ...  \n",
       "\n",
       "[5594 rows x 5 columns]"
      ]
     },
     "execution_count": 108,
     "metadata": {
      "nterop": {
       "id": "166"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_motion[df_motion.imdb_id.isin(split_ids['train'])]"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "62"
    }
   },
   "source": [
    "#### append split to each row"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {
    "nterop": {
     "id": "63"
    }
   },
   "outputs": [],
   "source": [
    "id_to_split = {imdb_id: s for s, ids in split_ids.items() for imdb_id in ids}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {
    "nterop": {
     "id": "64"
    }
   },
   "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>imdb_id</th>\n",
       "      <th>shot1_idx</th>\n",
       "      <th>shot2_idx</th>\n",
       "      <th>label</th>\n",
       "      <th>source_info</th>\n",
       "      <th>split</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>92</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>112</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>294</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>500</td>\n",
       "      <td>True</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>6</td>\n",
       "      <td>530</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 3: ...</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9315</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1203</td>\n",
       "      <td>1426</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9316</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1206</td>\n",
       "      <td>1367</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9317</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1505</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9318</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1558</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9319</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1721</td>\n",
       "      <td>False</td>\n",
       "      <td>SourceInfo(sources_and_scores=(('heuristic 4: ...</td>\n",
       "      <td>train</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>9320 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        imdb_id  shot1_idx  shot2_idx  label  \\\n",
       "0     tt0118799          6         92  False   \n",
       "1     tt0118799          6        112  False   \n",
       "2     tt0118799          6        294  False   \n",
       "3     tt0118799          6        500   True   \n",
       "4     tt0118799          6        530  False   \n",
       "...         ...        ...        ...    ...   \n",
       "9315  tt6788942       1203       1426  False   \n",
       "9316  tt6788942       1206       1367  False   \n",
       "9317  tt6788942       1367       1505  False   \n",
       "9318  tt6788942       1367       1558  False   \n",
       "9319  tt6788942       1367       1721  False   \n",
       "\n",
       "                                            source_info  split  \n",
       "0     SourceInfo(sources_and_scores=(('heuristic 3: ...   test  \n",
       "1     SourceInfo(sources_and_scores=(('heuristic 3: ...   test  \n",
       "2     SourceInfo(sources_and_scores=(('heuristic 3: ...   test  \n",
       "3     SourceInfo(sources_and_scores=(('heuristic 3: ...   test  \n",
       "4     SourceInfo(sources_and_scores=(('heuristic 3: ...   test  \n",
       "...                                                 ...    ...  \n",
       "9315  SourceInfo(sources_and_scores=(('heuristic 4: ...  train  \n",
       "9316  SourceInfo(sources_and_scores=(('heuristic 4: ...  train  \n",
       "9317  SourceInfo(sources_and_scores=(('heuristic 4: ...  train  \n",
       "9318  SourceInfo(sources_and_scores=(('heuristic 4: ...  train  \n",
       "9319  SourceInfo(sources_and_scores=(('heuristic 4: ...  train  \n",
       "\n",
       "[9320 rows x 6 columns]"
      ]
     },
     "execution_count": 110,
     "metadata": {
      "nterop": {
       "id": "167"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_motion.assign(split=df_motion.imdb_id.map(id_to_split))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "114"
    }
   },
   "source": [
    "### Annotator-level data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {
    "nterop": {
     "id": "115"
    }
   },
   "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>imdb_id</th>\n",
       "      <th>shot1_idx</th>\n",
       "      <th>shot2_idx</th>\n",
       "      <th>annotator_idx</th>\n",
       "      <th>annotator_label</th>\n",
       "      <th>task</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>3</td>\n",
       "      <td>13</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td>frame</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>3</td>\n",
       "      <td>13</td>\n",
       "      <td>2</td>\n",
       "      <td>False</td>\n",
       "      <td>frame</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>3</td>\n",
       "      <td>13</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>frame</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>3</td>\n",
       "      <td>14</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td>frame</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tt0118799</td>\n",
       "      <td>3</td>\n",
       "      <td>14</td>\n",
       "      <td>2</td>\n",
       "      <td>False</td>\n",
       "      <td>frame</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27955</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1558</td>\n",
       "      <td>2</td>\n",
       "      <td>False</td>\n",
       "      <td>motion</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27956</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1558</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>motion</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27957</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1721</td>\n",
       "      <td>1</td>\n",
       "      <td>False</td>\n",
       "      <td>motion</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27958</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1721</td>\n",
       "      <td>2</td>\n",
       "      <td>False</td>\n",
       "      <td>motion</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27959</th>\n",
       "      <td>tt6788942</td>\n",
       "      <td>1367</td>\n",
       "      <td>1721</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>motion</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>57915 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         imdb_id  shot1_idx  shot2_idx  annotator_idx  annotator_label    task\n",
       "0      tt0118799          3         13              1            False   frame\n",
       "1      tt0118799          3         13              2            False   frame\n",
       "2      tt0118799          3         13              3            False   frame\n",
       "3      tt0118799          3         14              1            False   frame\n",
       "4      tt0118799          3         14              2            False   frame\n",
       "...          ...        ...        ...            ...              ...     ...\n",
       "27955  tt6788942       1367       1558              2            False  motion\n",
       "27956  tt6788942       1367       1558              3            False  motion\n",
       "27957  tt6788942       1367       1721              1            False  motion\n",
       "27958  tt6788942       1367       1721              2            False  motion\n",
       "27959  tt6788942       1367       1721              3            False  motion\n",
       "\n",
       "[57915 rows x 6 columns]"
      ]
     },
     "execution_count": 133,
     "metadata": {
      "nterop": {
       "id": "168"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def get_annotator_level_labels(task: str) -> pd.DataFrame:\n",
    "    pairs = json.load(open(f\"data/dataset-{task}.json\"))\n",
    "    return pd.DataFrame(\n",
    "        dict(\n",
    "            imdb_id=pair['imdb_id'],\n",
    "            shot1_idx=pair['clip1_index'],\n",
    "            shot2_idx=pair['clip2_index'],\n",
    "            annotator_idx=int(idx),\n",
    "            annotator_label=label,\n",
    "            task=task,\n",
    "        )\n",
    "        for pair in pairs\n",
    "        for idx, label in pair[\"annotator_labels\"].items()\n",
    "    )\n",
    "\n",
    "df_annotator_level = pd.concat((\n",
    "    get_annotator_level_labels(task=\"frame\"),\n",
    "    get_annotator_level_labels(task=\"motion\"),\n",
    "))\n",
    "df_annotator_level"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {
    "nterop": {
     "id": "117"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "task\n",
       "frame     0.104991\n",
       "motion    0.131545\n",
       "Name: annotator_label, dtype: float64"
      ]
     },
     "execution_count": 134,
     "metadata": {
      "nterop": {
       "id": "169"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_annotator_level.groupby('task').annotator_label.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {
    "nterop": {
     "id": "119"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.11781058447725114"
      ]
     },
     "execution_count": 135,
     "metadata": {
      "nterop": {
       "id": "170"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_annotator_level.annotator_label.mean()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "121"
    }
   },
   "source": [
    "### Annotator agreement"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "nterop": {
     "id": "122"
    }
   },
   "outputs": [],
   "source": [
    "df_annotator_pos_cnt = (\n",
    "    df_annotator_level\n",
    "    .groupby(['task', 'imdb_id', 'shot1_idx', 'shot2_idx'])\n",
    "    .annotator_label\n",
    "    .sum()\n",
    "    .apply(int)\n",
    "    .rename('positive_label_cnt').reset_index()\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "nterop": {
     "id": "124"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "nterop": {
       "id": "171"
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "_ = df_annotator_pos_cnt.groupby(['positive_label_cnt', 'task']).size().unstack().plot(kind='bar', rot=0)\n",
    "plt.ylabel('Shot pairs')\n",
    "_ = plt.xlabel('Number of annotators that labeled the same pair positive')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "66"
    }
   },
   "source": [
    "### experiment results"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "67"
    }
   },
   "source": [
    "#### experiment 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {
    "nterop": {
     "id": "68"
    }
   },
   "outputs": [],
   "source": [
    "df_exp1 = pd.read_csv('experiment1.csv')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "174"
    }
   },
   "source": [
    "#### detailed view"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {
    "nterop": {
     "id": "175"
    }
   },
   "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>task</th>\n",
       "      <th>encoder</th>\n",
       "      <th>model</th>\n",
       "      <th>agg</th>\n",
       "      <th>train-AP</th>\n",
       "      <th>validation-AP</th>\n",
       "      <th>test-AP</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>frame</td>\n",
       "      <td>baseline</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.079</td>\n",
       "      <td>0.094</td>\n",
       "      <td>0.119</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>motion</td>\n",
       "      <td>baseline</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.095</td>\n",
       "      <td>0.096</td>\n",
       "      <td>0.122</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>frame</td>\n",
       "      <td>heuristic 1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.012</td>\n",
       "      <td>0.006</td>\n",
       "      <td>0.017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>frame</td>\n",
       "      <td>heuristic 2</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.169</td>\n",
       "      <td>0.177</td>\n",
       "      <td>0.207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>frame</td>\n",
       "      <td>heuristic 3</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.218</td>\n",
       "      <td>0.234</td>\n",
       "      <td>0.248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>392</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>mlp_small</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.778 ± 0.018</td>\n",
       "      <td>0.12 ± 0.006</td>\n",
       "      <td>0.167</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>393</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>mlp_small</td>\n",
       "      <td>abs_diff</td>\n",
       "      <td>0.773 ± 0.015</td>\n",
       "      <td>0.118 ± 0.008</td>\n",
       "      <td>0.132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>394</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.793 ± 0.022</td>\n",
       "      <td>0.121 ± 0.008</td>\n",
       "      <td>0.182</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>395</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.778 ± 0.014</td>\n",
       "      <td>0.122 ± 0.004</td>\n",
       "      <td>0.176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>396</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>abs_diff</td>\n",
       "      <td>0.786 ± 0.012</td>\n",
       "      <td>0.117 ± 0.012</td>\n",
       "      <td>0.138</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>397 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       task      encoder      model       agg       train-AP  validation-AP  \\\n",
       "0     frame     baseline        NaN       NaN          0.079          0.094   \n",
       "1    motion     baseline        NaN       NaN          0.095          0.096   \n",
       "2     frame  heuristic 1        NaN       NaN          0.012          0.006   \n",
       "3     frame  heuristic 2        NaN       NaN          0.169          0.177   \n",
       "4     frame  heuristic 3        NaN       NaN          0.218          0.234   \n",
       "..      ...          ...        ...       ...            ...            ...   \n",
       "392  motion  yamnet-swin  mlp_small    concat  0.778 ± 0.018   0.12 ± 0.006   \n",
       "393  motion  yamnet-swin  mlp_small  abs_diff  0.773 ± 0.015  0.118 ± 0.008   \n",
       "394  motion  yamnet-swin  mlp_large      mean  0.793 ± 0.022  0.121 ± 0.008   \n",
       "395  motion  yamnet-swin  mlp_large    concat  0.778 ± 0.014  0.122 ± 0.004   \n",
       "396  motion  yamnet-swin  mlp_large  abs_diff  0.786 ± 0.012  0.117 ± 0.012   \n",
       "\n",
       "     test-AP  \n",
       "0      0.119  \n",
       "1      0.122  \n",
       "2      0.017  \n",
       "3      0.207  \n",
       "4      0.248  \n",
       "..       ...  \n",
       "392    0.167  \n",
       "393    0.132  \n",
       "394    0.182  \n",
       "395    0.176  \n",
       "396    0.138  \n",
       "\n",
       "[397 rows x 7 columns]"
      ]
     },
     "execution_count": 168,
     "metadata": {
      "nterop": {
       "id": "176"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_exp1"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "177"
    }
   },
   "source": [
    "#### aggregated"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {
    "nterop": {
     "id": "178"
    }
   },
   "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>task</th>\n",
       "      <th>encoder</th>\n",
       "      <th>model</th>\n",
       "      <th>agg</th>\n",
       "      <th>train-AP</th>\n",
       "      <th>validation-AP</th>\n",
       "      <th>test-AP</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>frame</td>\n",
       "      <td>baseline</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.079</td>\n",
       "      <td>0.094</td>\n",
       "      <td>0.119</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>frame</td>\n",
       "      <td>clip</td>\n",
       "      <td>mlp</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.829 ± 0.011</td>\n",
       "      <td>0.253 ± 0.023</td>\n",
       "      <td>0.240</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>frame</td>\n",
       "      <td>clip4clip</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.836 ± 0.009</td>\n",
       "      <td>0.277 ± 0.012</td>\n",
       "      <td>0.304</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>frame</td>\n",
       "      <td>efficientnetb7</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.852 ± 0.011</td>\n",
       "      <td>0.261 ± 0.025</td>\n",
       "      <td>0.352</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>frame</td>\n",
       "      <td>heuristic 1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.012</td>\n",
       "      <td>0.006</td>\n",
       "      <td>0.017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>frame</td>\n",
       "      <td>heuristic 2</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.169</td>\n",
       "      <td>0.177</td>\n",
       "      <td>0.207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>frame</td>\n",
       "      <td>heuristic 3</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.218</td>\n",
       "      <td>0.234</td>\n",
       "      <td>0.248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>frame</td>\n",
       "      <td>r2plus1d</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.829 ± 0.012</td>\n",
       "      <td>0.222 ± 0.021</td>\n",
       "      <td>0.224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>frame</td>\n",
       "      <td>resnet50</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.838 ± 0.01</td>\n",
       "      <td>0.266 ± 0.015</td>\n",
       "      <td>0.269</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>frame</td>\n",
       "      <td>swin</td>\n",
       "      <td>mlp</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.815 ± 0.014</td>\n",
       "      <td>0.27 ± 0.03</td>\n",
       "      <td>0.287</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.81 ± 0.01</td>\n",
       "      <td>0.174 ± 0.015</td>\n",
       "      <td>0.160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet-clip</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.857 ± 0.008</td>\n",
       "      <td>0.245 ± 0.022</td>\n",
       "      <td>0.286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet-clip4clip</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.861 ± 0.009</td>\n",
       "      <td>0.289 ± 0.017</td>\n",
       "      <td>0.327</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet-efficientnetb7</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.856 ± 0.008</td>\n",
       "      <td>0.275 ± 0.03</td>\n",
       "      <td>0.355</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet-r2plus1d</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.836 ± 0.003</td>\n",
       "      <td>0.219 ± 0.02</td>\n",
       "      <td>0.218</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet-resnet50</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.85 ± 0.006</td>\n",
       "      <td>0.283 ± 0.015</td>\n",
       "      <td>0.321</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.81 ± 0.015</td>\n",
       "      <td>0.269 ± 0.024</td>\n",
       "      <td>0.336</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>motion</td>\n",
       "      <td>baseline</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.095</td>\n",
       "      <td>0.096</td>\n",
       "      <td>0.122</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>motion</td>\n",
       "      <td>clip</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.787 ± 0.021</td>\n",
       "      <td>0.131 ± 0.01</td>\n",
       "      <td>0.107</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>motion</td>\n",
       "      <td>clip4clip</td>\n",
       "      <td>mlp</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.8 ± 0.013</td>\n",
       "      <td>0.121 ± 0.007</td>\n",
       "      <td>0.130</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>motion</td>\n",
       "      <td>efficientnetb7</td>\n",
       "      <td>mlp_small</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.789 ± 0.012</td>\n",
       "      <td>0.118 ± 0.012</td>\n",
       "      <td>0.112</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>motion</td>\n",
       "      <td>heuristic 4</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.195</td>\n",
       "      <td>0.192</td>\n",
       "      <td>0.163</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>motion</td>\n",
       "      <td>heuristic 5</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.184</td>\n",
       "      <td>0.134</td>\n",
       "      <td>0.132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>motion</td>\n",
       "      <td>r2plus1d</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>concat</td>\n",
       "      <td>0.805 ± 0.009</td>\n",
       "      <td>0.184 ± 0.022</td>\n",
       "      <td>0.193</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>motion</td>\n",
       "      <td>resnet50</td>\n",
       "      <td>mlp</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.791 ± 0.016</td>\n",
       "      <td>0.12 ± 0.01</td>\n",
       "      <td>0.136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>motion</td>\n",
       "      <td>swin</td>\n",
       "      <td>mlp_small</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.787 ± 0.025</td>\n",
       "      <td>0.155 ± 0.016</td>\n",
       "      <td>0.150</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet</td>\n",
       "      <td>mlp_large</td>\n",
       "      <td>abs_diff</td>\n",
       "      <td>0.743 ± 0.019</td>\n",
       "      <td>0.136 ± 0.016</td>\n",
       "      <td>0.136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-clip</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>abs_diff</td>\n",
       "      <td>0.795 ± 0.009</td>\n",
       "      <td>0.137 ± 0.016</td>\n",
       "      <td>0.137</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-clip4clip</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.826 ± 0.013</td>\n",
       "      <td>0.14 ± 0.008</td>\n",
       "      <td>0.124</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-efficientnetb7</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>abs_diff</td>\n",
       "      <td>0.801 ± 0.011</td>\n",
       "      <td>0.136 ± 0.011</td>\n",
       "      <td>0.128</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-r2plus1d</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.835 ± 0.011</td>\n",
       "      <td>0.17 ± 0.015</td>\n",
       "      <td>0.177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-resnet50</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.821 ± 0.013</td>\n",
       "      <td>0.129 ± 0.007</td>\n",
       "      <td>0.145</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet-swin</td>\n",
       "      <td>xgboost</td>\n",
       "      <td>mean</td>\n",
       "      <td>0.824 ± 0.014</td>\n",
       "      <td>0.139 ± 0.01</td>\n",
       "      <td>0.161</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      task                encoder      model       agg       train-AP  \\\n",
       "0    frame               baseline        NaN       NaN          0.079   \n",
       "1    frame                   clip        mlp      mean  0.829 ± 0.011   \n",
       "2    frame              clip4clip  mlp_large      mean  0.836 ± 0.009   \n",
       "3    frame         efficientnetb7    xgboost      mean  0.852 ± 0.011   \n",
       "4    frame            heuristic 1        NaN       NaN          0.012   \n",
       "5    frame            heuristic 2        NaN       NaN          0.169   \n",
       "6    frame            heuristic 3        NaN       NaN          0.218   \n",
       "7    frame               r2plus1d  mlp_large      mean  0.829 ± 0.012   \n",
       "8    frame               resnet50  mlp_large    concat   0.838 ± 0.01   \n",
       "9    frame                   swin        mlp      mean  0.815 ± 0.014   \n",
       "10   frame                 yamnet    xgboost    concat    0.81 ± 0.01   \n",
       "11   frame            yamnet-clip    xgboost    concat  0.857 ± 0.008   \n",
       "12   frame       yamnet-clip4clip    xgboost      mean  0.861 ± 0.009   \n",
       "13   frame  yamnet-efficientnetb7    xgboost      mean  0.856 ± 0.008   \n",
       "14   frame        yamnet-r2plus1d    xgboost    concat  0.836 ± 0.003   \n",
       "15   frame        yamnet-resnet50    xgboost    concat   0.85 ± 0.006   \n",
       "16   frame            yamnet-swin  mlp_large      mean   0.81 ± 0.015   \n",
       "17  motion               baseline        NaN       NaN          0.095   \n",
       "18  motion                   clip  mlp_large    concat  0.787 ± 0.021   \n",
       "19  motion              clip4clip        mlp      mean    0.8 ± 0.013   \n",
       "20  motion         efficientnetb7  mlp_small      mean  0.789 ± 0.012   \n",
       "21  motion            heuristic 4        NaN       NaN          0.195   \n",
       "22  motion            heuristic 5        NaN       NaN          0.184   \n",
       "23  motion               r2plus1d  mlp_large    concat  0.805 ± 0.009   \n",
       "24  motion               resnet50        mlp      mean  0.791 ± 0.016   \n",
       "25  motion                   swin  mlp_small      mean  0.787 ± 0.025   \n",
       "26  motion                 yamnet  mlp_large  abs_diff  0.743 ± 0.019   \n",
       "27  motion            yamnet-clip    xgboost  abs_diff  0.795 ± 0.009   \n",
       "28  motion       yamnet-clip4clip    xgboost      mean  0.826 ± 0.013   \n",
       "29  motion  yamnet-efficientnetb7    xgboost  abs_diff  0.801 ± 0.011   \n",
       "30  motion        yamnet-r2plus1d    xgboost      mean  0.835 ± 0.011   \n",
       "31  motion        yamnet-resnet50    xgboost      mean  0.821 ± 0.013   \n",
       "32  motion            yamnet-swin    xgboost      mean  0.824 ± 0.014   \n",
       "\n",
       "    validation-AP  test-AP  \n",
       "0           0.094    0.119  \n",
       "1   0.253 ± 0.023    0.240  \n",
       "2   0.277 ± 0.012    0.304  \n",
       "3   0.261 ± 0.025    0.352  \n",
       "4           0.006    0.017  \n",
       "5           0.177    0.207  \n",
       "6           0.234    0.248  \n",
       "7   0.222 ± 0.021    0.224  \n",
       "8   0.266 ± 0.015    0.269  \n",
       "9     0.27 ± 0.03    0.287  \n",
       "10  0.174 ± 0.015    0.160  \n",
       "11  0.245 ± 0.022    0.286  \n",
       "12  0.289 ± 0.017    0.327  \n",
       "13   0.275 ± 0.03    0.355  \n",
       "14   0.219 ± 0.02    0.218  \n",
       "15  0.283 ± 0.015    0.321  \n",
       "16  0.269 ± 0.024    0.336  \n",
       "17          0.096    0.122  \n",
       "18   0.131 ± 0.01    0.107  \n",
       "19  0.121 ± 0.007    0.130  \n",
       "20  0.118 ± 0.012    0.112  \n",
       "21          0.192    0.163  \n",
       "22          0.134    0.132  \n",
       "23  0.184 ± 0.022    0.193  \n",
       "24    0.12 ± 0.01    0.136  \n",
       "25  0.155 ± 0.016    0.150  \n",
       "26  0.136 ± 0.016    0.136  \n",
       "27  0.137 ± 0.016    0.137  \n",
       "28   0.14 ± 0.008    0.124  \n",
       "29  0.136 ± 0.011    0.128  \n",
       "30   0.17 ± 0.015    0.177  \n",
       "31  0.129 ± 0.007    0.145  \n",
       "32   0.139 ± 0.01    0.161  "
      ]
     },
     "execution_count": 171,
     "metadata": {
      "nterop": {
       "id": "179"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(\n",
    "    df_exp1\n",
    "    .assign(val_mean=df_exp1['validation-AP'].apply(lambda x: float(x.split(' ')[0])))\n",
    "    .groupby(['task', 'encoder'])\n",
    "    # take the combination with the highest mean validation AP per (task, encoder)\n",
    "    .apply(lambda rows: rows.sort_values(by='val_mean', ascending=False).head(1))\n",
    "    .drop(columns=['val_mean'])\n",
    "    .reset_index(drop=True)\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "nterop": {
     "id": "180"
    }
   },
   "source": [
    "### experiment 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {
    "nterop": {
     "id": "70"
    }
   },
   "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>task</th>\n",
       "      <th>encoder</th>\n",
       "      <th>model</th>\n",
       "      <th>agg</th>\n",
       "      <th>train-AP</th>\n",
       "      <th>validation-AP</th>\n",
       "      <th>test-AP</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>frame</td>\n",
       "      <td>clip4clip</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.751 ± 0.005</td>\n",
       "      <td>0.273 ± 0.01</td>\n",
       "      <td>0.359</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>frame</td>\n",
       "      <td>clip</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.743 ± 0.005</td>\n",
       "      <td>0.232 ± 0.009</td>\n",
       "      <td>0.320</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>frame</td>\n",
       "      <td>r2plus1d</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.768 ± 0.004</td>\n",
       "      <td>0.218 ± 0.007</td>\n",
       "      <td>0.263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>frame</td>\n",
       "      <td>efficientnetb7</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.778 ± 0.004</td>\n",
       "      <td>0.284 ± 0.01</td>\n",
       "      <td>0.373</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>frame</td>\n",
       "      <td>resnet50</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.771 ± 0.006</td>\n",
       "      <td>0.258 ± 0.01</td>\n",
       "      <td>0.342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>frame</td>\n",
       "      <td>swin</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.75 ± 0.004</td>\n",
       "      <td>0.255 ± 0.005</td>\n",
       "      <td>0.361</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>frame</td>\n",
       "      <td>yamnet</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.355 ± 0.016</td>\n",
       "      <td>0.118 ± 0.003</td>\n",
       "      <td>0.097</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>motion</td>\n",
       "      <td>r2plus1d</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.713 ± 0.007</td>\n",
       "      <td>0.173 ± 0.001</td>\n",
       "      <td>0.217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>motion</td>\n",
       "      <td>swin</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.702 ± 0.006</td>\n",
       "      <td>0.143 ± 0.002</td>\n",
       "      <td>0.170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>motion</td>\n",
       "      <td>clip4clip</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.704 ± 0.006</td>\n",
       "      <td>0.119 ± 0.001</td>\n",
       "      <td>0.144</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>motion</td>\n",
       "      <td>resnet50</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.734 ± 0.003</td>\n",
       "      <td>0.119 ± 0.003</td>\n",
       "      <td>0.131</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>motion</td>\n",
       "      <td>efficientnetb7</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.75 ± 0.006</td>\n",
       "      <td>0.139 ± 0.004</td>\n",
       "      <td>0.132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>motion</td>\n",
       "      <td>clip</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.699 ± 0.006</td>\n",
       "      <td>0.112 ± 0.001</td>\n",
       "      <td>0.138</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>motion</td>\n",
       "      <td>yamnet</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.475 ± 0.009</td>\n",
       "      <td>0.122 ± 0.001</td>\n",
       "      <td>0.123</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      task         encoder  model  agg       train-AP  validation-AP  test-AP\n",
       "0    frame       clip4clip    NaN  NaN  0.751 ± 0.005   0.273 ± 0.01    0.359\n",
       "1    frame            clip    NaN  NaN  0.743 ± 0.005  0.232 ± 0.009    0.320\n",
       "2    frame        r2plus1d    NaN  NaN  0.768 ± 0.004  0.218 ± 0.007    0.263\n",
       "3    frame  efficientnetb7    NaN  NaN  0.778 ± 0.004   0.284 ± 0.01    0.373\n",
       "4    frame        resnet50    NaN  NaN  0.771 ± 0.006   0.258 ± 0.01    0.342\n",
       "5    frame            swin    NaN  NaN   0.75 ± 0.004  0.255 ± 0.005    0.361\n",
       "6    frame          yamnet    NaN  NaN  0.355 ± 0.016  0.118 ± 0.003    0.097\n",
       "7   motion        r2plus1d    NaN  NaN  0.713 ± 0.007  0.173 ± 0.001    0.217\n",
       "8   motion            swin    NaN  NaN  0.702 ± 0.006  0.143 ± 0.002    0.170\n",
       "9   motion       clip4clip    NaN  NaN  0.704 ± 0.006  0.119 ± 0.001    0.144\n",
       "10  motion        resnet50    NaN  NaN  0.734 ± 0.003  0.119 ± 0.003    0.131\n",
       "11  motion  efficientnetb7    NaN  NaN   0.75 ± 0.006  0.139 ± 0.004    0.132\n",
       "12  motion            clip    NaN  NaN  0.699 ± 0.006  0.112 ± 0.001    0.138\n",
       "13  motion          yamnet    NaN  NaN  0.475 ± 0.009  0.122 ± 0.001    0.123"
      ]
     },
     "execution_count": 156,
     "metadata": {
      "nterop": {
       "id": "181"
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.read_csv('experiment2.csv')"
   ]
  }
 ],
 "metadata": {
  "hide_input": false,
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  },
  "nterop": {
   "seedId": "181"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}