{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "\n",
    "sns.set_style(\"whitegrid\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Reading in our potential test sets from [here](https://huggingface.co/datasets/imageomics/lila-bc-camera/tree/d18307b285217d18b31d1a7b2c9091bb0873ade0/data/potential-test-sets):\n",
    "  - [Ohio Small Animals](https://lila.science/datasets/ohio-small-animals/)\n",
    "  - [Desert Lion Conservation Camera Traps](https://lila.science/datasets/desert-lion-conservation-camera-traps/)\n",
    "  - [Orinoquia Camera Traps](https://lila.science/datasets/orinoquia-camera-traps/)\n",
    "  - [Island Conservation Camera Traps](https://lila.science/datasets/island-conservation-camera-traps/)\n",
    "  - [ENA24](https://lila.science/datasets/ena24detection)\n",
    "\n",
    "We'll clean them down to just the taxa and identifier columns, then further reduce to make balanced test sets for each.\n",
    "\n",
    "# ENA24 Datasets\n",
    "upper/lower bounded and balanced"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dataset_name</th>\n",
       "      <th>url_gcp</th>\n",
       "      <th>url_aws</th>\n",
       "      <th>url_azure</th>\n",
       "      <th>image_id</th>\n",
       "      <th>sequence_id</th>\n",
       "      <th>location_id</th>\n",
       "      <th>frame_num</th>\n",
       "      <th>original_label</th>\n",
       "      <th>scientific_name</th>\n",
       "      <th>...</th>\n",
       "      <th>superfamily</th>\n",
       "      <th>family</th>\n",
       "      <th>subfamily</th>\n",
       "      <th>tribe</th>\n",
       "      <th>genus</th>\n",
       "      <th>species</th>\n",
       "      <th>subspecies</th>\n",
       "      <th>variety</th>\n",
       "      <th>multi_species</th>\n",
       "      <th>num_species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>bird</td>\n",
       "      <td>aves</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 10</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>bird</td>\n",
       "      <td>aves</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 100</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern gray squirrel</td>\n",
       "      <td>sciurus carolinensis</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>sciurinae</td>\n",
       "      <td>sciurini</td>\n",
       "      <td>sciurus</td>\n",
       "      <td>sciurus carolinensis</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1000</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>xerinae</td>\n",
       "      <td>tamiini</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1001</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>xerinae</td>\n",
       "      <td>tamiini</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 34 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  dataset_name                                            url_gcp  \\\n",
       "0        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "1        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "2        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "3        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "4        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "\n",
       "                                             url_aws  \\\n",
       "0  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "1  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "2  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "3  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "4  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "\n",
       "                                           url_azure      image_id  \\\n",
       "0  https://lilawildlife.blob.core.windows.net/lil...     ENA24 : 1   \n",
       "1  https://lilawildlife.blob.core.windows.net/lil...    ENA24 : 10   \n",
       "2  https://lilawildlife.blob.core.windows.net/lil...   ENA24 : 100   \n",
       "3  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1000   \n",
       "4  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1001   \n",
       "\n",
       "       sequence_id location_id  frame_num         original_label  \\\n",
       "0  ENA24 : unknown       ENA24         -1                   bird   \n",
       "1  ENA24 : unknown       ENA24         -1                   bird   \n",
       "2  ENA24 : unknown       ENA24         -1  eastern gray squirrel   \n",
       "3  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "4  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "\n",
       "        scientific_name  ... superfamily     family  subfamily     tribe  \\\n",
       "0                  aves  ...         NaN        NaN        NaN       NaN   \n",
       "1                  aves  ...         NaN        NaN        NaN       NaN   \n",
       "2  sciurus carolinensis  ...         NaN  sciuridae  sciurinae  sciurini   \n",
       "3       tamias striatus  ...         NaN  sciuridae    xerinae   tamiini   \n",
       "4       tamias striatus  ...         NaN  sciuridae    xerinae   tamiini   \n",
       "\n",
       "     genus               species  subspecies variety multi_species num_species  \n",
       "0      NaN                   NaN         NaN     NaN         False         1.0  \n",
       "1      NaN                   NaN         NaN     NaN         False         1.0  \n",
       "2  sciurus  sciurus carolinensis         NaN     NaN         False         1.0  \n",
       "3   tamias       tamias striatus         NaN     NaN         False         1.0  \n",
       "4   tamias       tamias striatus         NaN     NaN         False         1.0  \n",
       "\n",
       "[5 rows x 34 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"../data/potential-test-sets/ENA24_image_urls_and_labels.csv\", low_memory = False)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['dataset_name', 'url_gcp', 'url_aws', 'url_azure', 'image_id',\n",
       "       'sequence_id', 'location_id', 'frame_num', 'original_label',\n",
       "       'scientific_name', 'common_name', 'datetime', 'annotation_level',\n",
       "       'kingdom', 'phylum', 'subphylum', 'superclass', 'class', 'subclass',\n",
       "       'infraclass', 'superorder', 'order', 'suborder', 'infraorder',\n",
       "       'superfamily', 'family', 'subfamily', 'tribe', 'genus', 'species',\n",
       "       'subspecies', 'variety', 'multi_species', 'num_species'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Observe that we also now get multiple URL options; `url_aws` will likely be best/fastest for use with [`distributed-downloader`](https://github.com/Imageomics/distributed-downloader) to get the images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 8758 entries, 0 to 8757\n",
      "Data columns (total 34 columns):\n",
      " #   Column            Non-Null Count  Dtype  \n",
      "---  ------            --------------  -----  \n",
      " 0   dataset_name      8758 non-null   object \n",
      " 1   url_gcp           8758 non-null   object \n",
      " 2   url_aws           8758 non-null   object \n",
      " 3   url_azure         8758 non-null   object \n",
      " 4   image_id          8758 non-null   object \n",
      " 5   sequence_id       8758 non-null   object \n",
      " 6   location_id       8758 non-null   object \n",
      " 7   frame_num         8758 non-null   int64  \n",
      " 8   original_label    8758 non-null   object \n",
      " 9   scientific_name   8758 non-null   object \n",
      " 10  common_name       8758 non-null   object \n",
      " 11  datetime          0 non-null      float64\n",
      " 12  annotation_level  8758 non-null   object \n",
      " 13  kingdom           8758 non-null   object \n",
      " 14  phylum            8758 non-null   object \n",
      " 15  subphylum         8758 non-null   object \n",
      " 16  superclass        0 non-null      float64\n",
      " 17  class             8758 non-null   object \n",
      " 18  subclass          6791 non-null   object \n",
      " 19  infraclass        6791 non-null   object \n",
      " 20  superorder        6066 non-null   object \n",
      " 21  order             8558 non-null   object \n",
      " 22  suborder          1512 non-null   object \n",
      " 23  infraorder        0 non-null      float64\n",
      " 24  superfamily       0 non-null      float64\n",
      " 25  family            8558 non-null   object \n",
      " 26  subfamily         4237 non-null   object \n",
      " 27  tribe             2237 non-null   object \n",
      " 28  genus             8558 non-null   object \n",
      " 29  species           8558 non-null   object \n",
      " 30  subspecies        340 non-null    object \n",
      " 31  variety           531 non-null    object \n",
      " 32  multi_species     8758 non-null   bool   \n",
      " 33  num_species       8758 non-null   float64\n",
      "dtypes: bool(1), float64(5), int64(1), object(27)\n",
      "memory usage: 2.2+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info(show_counts = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This is mostly filled in--only 200 don't have species label (all go to class). Let's get some counts on the taxa, then start looking at balancing it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "lin_taxa = ['kingdom', 'phylum', 'class', 'order', 'family', 'genus', 'species']\n",
    "taxa_cols = ['original_label', 'scientific_name', 'common_name', 'kingdom', 'phylum', 'class', 'order', 'family', 'genus', 'species']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "original_label     21\n",
       "scientific_name    21\n",
       "common_name        21\n",
       "kingdom             1\n",
       "phylum              1\n",
       "class               2\n",
       "order               8\n",
       "family             12\n",
       "genus              18\n",
       "species            20\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[taxa_cols].nunique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We only have 21 scientific names, 20 species, so the difference should be the null species values.\n",
    "\n",
    "Also, should check for duplicated images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of unique images: 8652\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "multi_species\n",
       "False    8546\n",
       "True      212\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(f\"number of unique images: {df[\"image_id\"].nunique()}\")\n",
    "\n",
    "df[\"multi_species\"].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have 106 duplicated images (multiple species per image), so let's check on these."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "num_species\n",
       "1.0    8546\n",
       "2.0     212\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"num_species\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "scientific_name\n",
       "sylvilagus floridanus       76\n",
       "corvus brachyrhynchos       68\n",
       "gallus gallus domesticus    20\n",
       "odocoileus virginianus      17\n",
       "canis familiaris             7\n",
       "sciurus carolinensis         7\n",
       "didelphis virginiana         5\n",
       "equus caballus               4\n",
       "felis catus                  3\n",
       "lynx rufus                   3\n",
       "canis latrans                2\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.loc[df[\"multi_species\"], \"scientific_name\"].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "All of them are just 2 species, let's check if we lose anything (for record), then filter."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "21"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.loc[~df[\"multi_species\"], \"scientific_name\"].nunique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Okay, none are dropped, so now let's filter then look at how many images we have per species."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "image_id           8546\n",
       "scientific_name      21\n",
       "class                 2\n",
       "species              20\n",
       "dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_single = df.loc[~df[\"multi_species\"]].copy()\n",
    "print(df[\"class\"].nunique())\n",
    "df_single[[\"image_id\",\"scientific_name\", \"class\", \"species\"]].nunique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dataset_name</th>\n",
       "      <th>url_gcp</th>\n",
       "      <th>url_aws</th>\n",
       "      <th>url_azure</th>\n",
       "      <th>image_id</th>\n",
       "      <th>sequence_id</th>\n",
       "      <th>location_id</th>\n",
       "      <th>frame_num</th>\n",
       "      <th>original_label</th>\n",
       "      <th>scientific_name</th>\n",
       "      <th>...</th>\n",
       "      <th>superfamily</th>\n",
       "      <th>family</th>\n",
       "      <th>subfamily</th>\n",
       "      <th>tribe</th>\n",
       "      <th>genus</th>\n",
       "      <th>species</th>\n",
       "      <th>subspecies</th>\n",
       "      <th>variety</th>\n",
       "      <th>multi_species</th>\n",
       "      <th>num_species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>699</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1628</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>bird</td>\n",
       "      <td>aves</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>706</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1634</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>bird</td>\n",
       "      <td>aves</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>720</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1647</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>bird</td>\n",
       "      <td>aves</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2720</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 37</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>bird</td>\n",
       "      <td>aves</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4 rows × 34 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     dataset_name                                            url_gcp  \\\n",
       "699         ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "706         ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "720         ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "2720        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "\n",
       "                                                url_aws  \\\n",
       "699   http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "706   http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "720   http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "2720  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "\n",
       "                                              url_azure      image_id  \\\n",
       "699   https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1628   \n",
       "706   https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1634   \n",
       "720   https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1647   \n",
       "2720  https://lilawildlife.blob.core.windows.net/lil...    ENA24 : 37   \n",
       "\n",
       "          sequence_id location_id  frame_num original_label scientific_name  \\\n",
       "699   ENA24 : unknown       ENA24         -1           bird            aves   \n",
       "706   ENA24 : unknown       ENA24         -1           bird            aves   \n",
       "720   ENA24 : unknown       ENA24         -1           bird            aves   \n",
       "2720  ENA24 : unknown       ENA24         -1           bird            aves   \n",
       "\n",
       "      ... superfamily  family subfamily tribe genus species  subspecies  \\\n",
       "699   ...         NaN     NaN       NaN   NaN   NaN     NaN         NaN   \n",
       "706   ...         NaN     NaN       NaN   NaN   NaN     NaN         NaN   \n",
       "720   ...         NaN     NaN       NaN   NaN   NaN     NaN         NaN   \n",
       "2720  ...         NaN     NaN       NaN   NaN   NaN     NaN         NaN   \n",
       "\n",
       "     variety multi_species num_species  \n",
       "699      NaN         False         1.0  \n",
       "706      NaN         False         1.0  \n",
       "720      NaN         False         1.0  \n",
       "2720     NaN         False         1.0  \n",
       "\n",
       "[4 rows x 34 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_single.loc[df_single[\"species\"].isna()].sample(4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Yes, class level."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "scientific_name\n",
       "ursus americanus            893\n",
       "corvus brachyrhynchos       879\n",
       "didelphis virginiana        720\n",
       "canis familiaris            696\n",
       "gallus gallus domesticus    511\n",
       "felis catus                 479\n",
       "urocyon cinereoargenteus    422\n",
       "vulpes vulpes               413\n",
       "sciurus niger cinereus      340\n",
       "odocoileus virginianus      333\n",
       "canis latrans               332\n",
       "lynx rufus                  325\n",
       "tamias striatus             311\n",
       "sciurus carolinensis        298\n",
       "mephitis mephitis           297\n",
       "procyon lotor               291\n",
       "meleagris gallopavo         289\n",
       "sylvilagus floridanus       255\n",
       "marmota monax               206\n",
       "aves                        200\n",
       "equus caballus               56\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_single.scientific_name.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Least represented species has 56 images! This set also isn't highly imbalanced. Interesting, so we can balance at 56 (or maybe 50 for round number)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: xlabel='Count', ylabel='class'>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.histplot(df_single, y = 'class')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Filter null species classifications"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Index: 8346 entries, 2 to 8757\n",
      "Data columns (total 10 columns):\n",
      " #   Column           Non-Null Count  Dtype \n",
      "---  ------           --------------  ----- \n",
      " 0   original_label   8346 non-null   object\n",
      " 1   scientific_name  8346 non-null   object\n",
      " 2   common_name      8346 non-null   object\n",
      " 3   kingdom          8346 non-null   object\n",
      " 4   phylum           8346 non-null   object\n",
      " 5   class            8346 non-null   object\n",
      " 6   order            8346 non-null   object\n",
      " 7   family           8346 non-null   object\n",
      " 8   genus            8346 non-null   object\n",
      " 9   species          8346 non-null   object\n",
      "dtypes: object(10)\n",
      "memory usage: 717.2+ KB\n"
     ]
    }
   ],
   "source": [
    "df_filter = df_single.loc[df_single[\"species\"].notna()].copy()\n",
    "df_filter[taxa_cols].info(show_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_filter[\"subspecies\"].nunique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "subspecies\n",
       "sciurus niger cinereus    340\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_filter[\"subspecies\"].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can check if these species are in here without subspecies designations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_filter.loc[(df_filter[\"species\"] == \"sciurus niger\") & (df_filter[\"subspecies\"].isna())].shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Yes, they're the only way these species are represented, so we can remove the subspecies."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Remove extra columns\n",
    "\n",
    "Only need `taxa_cols` (Linnean taxonomy + `original_label`, `scientific_name`, and `common_name`) and `id_cols`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "id_cols = ['dataset_name',\n",
    "           'url_gcp',\n",
    "           'url_aws',\n",
    "           'url_azure',\n",
    "           'image_id',\n",
    "           'sequence_id',\n",
    "           'location_id',\n",
    "           'frame_num']\n",
    "\n",
    "cols_to_keep = [col for col in list(df.columns) if (col in id_cols or col in taxa_cols)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['dataset_name',\n",
       " 'url_gcp',\n",
       " 'url_aws',\n",
       " 'url_azure',\n",
       " 'image_id',\n",
       " 'sequence_id',\n",
       " 'location_id',\n",
       " 'frame_num',\n",
       " 'original_label',\n",
       " 'scientific_name',\n",
       " 'common_name',\n",
       " 'kingdom',\n",
       " 'phylum',\n",
       " 'class',\n",
       " 'order',\n",
       " 'family',\n",
       " 'genus',\n",
       " 'species']"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols_to_keep"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's add a number of images column (by `scientific_name`)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dataset_name</th>\n",
       "      <th>url_gcp</th>\n",
       "      <th>url_aws</th>\n",
       "      <th>url_azure</th>\n",
       "      <th>image_id</th>\n",
       "      <th>sequence_id</th>\n",
       "      <th>location_id</th>\n",
       "      <th>frame_num</th>\n",
       "      <th>original_label</th>\n",
       "      <th>scientific_name</th>\n",
       "      <th>...</th>\n",
       "      <th>family</th>\n",
       "      <th>subfamily</th>\n",
       "      <th>tribe</th>\n",
       "      <th>genus</th>\n",
       "      <th>species</th>\n",
       "      <th>subspecies</th>\n",
       "      <th>variety</th>\n",
       "      <th>multi_species</th>\n",
       "      <th>num_species</th>\n",
       "      <th>num_sp_images</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 100</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern gray squirrel</td>\n",
       "      <td>sciurus carolinensis</td>\n",
       "      <td>...</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>sciurinae</td>\n",
       "      <td>sciurini</td>\n",
       "      <td>sciurus</td>\n",
       "      <td>sciurus carolinensis</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "      <td>298.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1000</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>...</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>xerinae</td>\n",
       "      <td>tamiini</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1001</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>...</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>xerinae</td>\n",
       "      <td>tamiini</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1002</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>...</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>xerinae</td>\n",
       "      <td>tamiini</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1003</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>...</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>xerinae</td>\n",
       "      <td>tamiini</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>False</td>\n",
       "      <td>1.0</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 35 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  dataset_name                                            url_gcp  \\\n",
       "2        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "3        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "4        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "5        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "6        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "\n",
       "                                             url_aws  \\\n",
       "2  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "3  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "4  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "5  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "6  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "\n",
       "                                           url_azure      image_id  \\\n",
       "2  https://lilawildlife.blob.core.windows.net/lil...   ENA24 : 100   \n",
       "3  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1000   \n",
       "4  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1001   \n",
       "5  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1002   \n",
       "6  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1003   \n",
       "\n",
       "       sequence_id location_id  frame_num         original_label  \\\n",
       "2  ENA24 : unknown       ENA24         -1  eastern gray squirrel   \n",
       "3  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "4  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "5  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "6  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "\n",
       "        scientific_name  ...     family  subfamily     tribe    genus  \\\n",
       "2  sciurus carolinensis  ...  sciuridae  sciurinae  sciurini  sciurus   \n",
       "3       tamias striatus  ...  sciuridae    xerinae   tamiini   tamias   \n",
       "4       tamias striatus  ...  sciuridae    xerinae   tamiini   tamias   \n",
       "5       tamias striatus  ...  sciuridae    xerinae   tamiini   tamias   \n",
       "6       tamias striatus  ...  sciuridae    xerinae   tamiini   tamias   \n",
       "\n",
       "                species subspecies  variety multi_species num_species  \\\n",
       "2  sciurus carolinensis        NaN      NaN         False         1.0   \n",
       "3       tamias striatus        NaN      NaN         False         1.0   \n",
       "4       tamias striatus        NaN      NaN         False         1.0   \n",
       "5       tamias striatus        NaN      NaN         False         1.0   \n",
       "6       tamias striatus        NaN      NaN         False         1.0   \n",
       "\n",
       "  num_sp_images  \n",
       "2         298.0  \n",
       "3         311.0  \n",
       "4         311.0  \n",
       "5         311.0  \n",
       "6         311.0  \n",
       "\n",
       "[5 rows x 35 columns]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "for sci_name in list(df_filter[\"scientific_name\"].unique()):\n",
    "    df_filter.loc[df_filter[\"scientific_name\"] == sci_name, \"num_sp_images\"] = df_filter.loc[df_filter[\"scientific_name\"] == sci_name].shape[0]\n",
    "\n",
    "df_filter.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "cols_to_keep.append(\"num_sp_images\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dataset_name</th>\n",
       "      <th>url_gcp</th>\n",
       "      <th>url_aws</th>\n",
       "      <th>url_azure</th>\n",
       "      <th>image_id</th>\n",
       "      <th>sequence_id</th>\n",
       "      <th>location_id</th>\n",
       "      <th>frame_num</th>\n",
       "      <th>original_label</th>\n",
       "      <th>scientific_name</th>\n",
       "      <th>common_name</th>\n",
       "      <th>kingdom</th>\n",
       "      <th>phylum</th>\n",
       "      <th>class</th>\n",
       "      <th>order</th>\n",
       "      <th>family</th>\n",
       "      <th>genus</th>\n",
       "      <th>species</th>\n",
       "      <th>num_sp_images</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 100</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern gray squirrel</td>\n",
       "      <td>sciurus carolinensis</td>\n",
       "      <td>eastern gray squirrel</td>\n",
       "      <td>animalia</td>\n",
       "      <td>chordata</td>\n",
       "      <td>mammalia</td>\n",
       "      <td>rodentia</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>sciurus</td>\n",
       "      <td>sciurus carolinensis</td>\n",
       "      <td>298.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1000</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>animalia</td>\n",
       "      <td>chordata</td>\n",
       "      <td>mammalia</td>\n",
       "      <td>rodentia</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1001</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>animalia</td>\n",
       "      <td>chordata</td>\n",
       "      <td>mammalia</td>\n",
       "      <td>rodentia</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1002</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>animalia</td>\n",
       "      <td>chordata</td>\n",
       "      <td>mammalia</td>\n",
       "      <td>rodentia</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>ENA24</td>\n",
       "      <td>https://storage.googleapis.com/public-datasets...</td>\n",
       "      <td>http://us-west-2.opendata.source.coop.s3.amazo...</td>\n",
       "      <td>https://lilawildlife.blob.core.windows.net/lil...</td>\n",
       "      <td>ENA24 : 1003</td>\n",
       "      <td>ENA24 : unknown</td>\n",
       "      <td>ENA24</td>\n",
       "      <td>-1</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>eastern chipmunk</td>\n",
       "      <td>animalia</td>\n",
       "      <td>chordata</td>\n",
       "      <td>mammalia</td>\n",
       "      <td>rodentia</td>\n",
       "      <td>sciuridae</td>\n",
       "      <td>tamias</td>\n",
       "      <td>tamias striatus</td>\n",
       "      <td>311.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  dataset_name                                            url_gcp  \\\n",
       "2        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "3        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "4        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "5        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "6        ENA24  https://storage.googleapis.com/public-datasets...   \n",
       "\n",
       "                                             url_aws  \\\n",
       "2  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "3  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "4  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "5  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "6  http://us-west-2.opendata.source.coop.s3.amazo...   \n",
       "\n",
       "                                           url_azure      image_id  \\\n",
       "2  https://lilawildlife.blob.core.windows.net/lil...   ENA24 : 100   \n",
       "3  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1000   \n",
       "4  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1001   \n",
       "5  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1002   \n",
       "6  https://lilawildlife.blob.core.windows.net/lil...  ENA24 : 1003   \n",
       "\n",
       "       sequence_id location_id  frame_num         original_label  \\\n",
       "2  ENA24 : unknown       ENA24         -1  eastern gray squirrel   \n",
       "3  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "4  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "5  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "6  ENA24 : unknown       ENA24         -1       eastern chipmunk   \n",
       "\n",
       "        scientific_name            common_name   kingdom    phylum     class  \\\n",
       "2  sciurus carolinensis  eastern gray squirrel  animalia  chordata  mammalia   \n",
       "3       tamias striatus       eastern chipmunk  animalia  chordata  mammalia   \n",
       "4       tamias striatus       eastern chipmunk  animalia  chordata  mammalia   \n",
       "5       tamias striatus       eastern chipmunk  animalia  chordata  mammalia   \n",
       "6       tamias striatus       eastern chipmunk  animalia  chordata  mammalia   \n",
       "\n",
       "      order     family    genus               species  num_sp_images  \n",
       "2  rodentia  sciuridae  sciurus  sciurus carolinensis          298.0  \n",
       "3  rodentia  sciuridae   tamias       tamias striatus          311.0  \n",
       "4  rodentia  sciuridae   tamias       tamias striatus          311.0  \n",
       "5  rodentia  sciuridae   tamias       tamias striatus          311.0  \n",
       "6  rodentia  sciuridae   tamias       tamias striatus          311.0  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_reduced = df_filter[cols_to_keep].copy()\n",
    "df_reduced.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We have at most 893 images per species, so no reduction needed\n",
    "Also, least represented species has 56 images, so no lower bound either."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_reduced.to_csv(\"../data/potential-test-sets/filtered/ENA24-imbalanced.csv\", index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: xlabel='Count', ylabel='scientific_name'>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.histplot(df_reduced, y = \"scientific_name\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Randomly sample to balanced set (12 images per species)\n",
    "\n",
    "Consistent with other datasets. Will also make a 56 image one."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "240"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "balanced_set_small = []\n",
    "for sci_name in list(df_reduced[\"scientific_name\"].unique()):\n",
    "    temp = df_reduced.loc[df_reduced[\"scientific_name\"] == sci_name].copy()\n",
    "    if temp.shape[0] < 12:\n",
    "        continue\n",
    "    sample_set = list(temp.sample(12, random_state = 614)[\"image_id\"])\n",
    "    balanced_set_small = balanced_set_small + sample_set\n",
    "\n",
    "len(balanced_set_small)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Filter to just balanced set and drop the number of species column since it's been balanced to 12 each."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_balanced_small = df_reduced.loc[df_reduced[\"image_id\"].isin(balanced_set_small)].copy()\n",
    "df_balanced_small.drop(columns = [\"num_sp_images\"], inplace = True)\n",
    "df_balanced_small.to_csv(\"../data/potential-test-sets/filtered/ENA24-balanced-small.csv\", index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1120"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "balanced_set = []\n",
    "for sci_name in list(df_reduced[\"scientific_name\"].unique()):\n",
    "    temp = df_reduced.loc[df_reduced[\"scientific_name\"] == sci_name].copy()\n",
    "    sample_set = list(temp.sample(56, random_state = 614)[\"image_id\"])\n",
    "    balanced_set = balanced_set + sample_set\n",
    "\n",
    "len(balanced_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_balanced = df_reduced.loc[df_reduced[\"image_id\"].isin(balanced_set)].copy()\n",
    "df_balanced.drop(columns = [\"num_sp_images\"], inplace = True)\n",
    "df_balanced.to_csv(\"../data/potential-test-sets/filtered/ENA24-balanced.csv\", index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "jupytext": {
   "formats": "ipynb,py:percent"
  },
  "kernelspec": {
   "display_name": "data-dev",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}