Recreate Salavon's piece 'Every Playboy Centerfold, The 1970s (normalized)' from 2002

salavon-centerfolds-recreate.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "90f14a7f-5f5a-4c92-bc5d-2e779aecd7bf",
   "metadata": {},
   "source": [
    "# Recreate Salavon's piece 'Every Playboy Centerfold, The 1970s (normalized)' from 2002\n",
    "This notebook recreates [Salavon's piece](http://salavon.com/work/everyplayboycenterfolddecades/image/216/), although not faithfully.\n",
    "Please also see the slides of the talk that goes along with it:\n",
    "\n",
    "[https://s.flx.ai/2022/digital-art-forgery/](https://s.flx.ai/2022/digital-art-forgery/)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b69f4596-fb6b-4a7c-a329-7dac1ab26b3f",
   "metadata": {},
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4799c2da-4c24-4cb1-abe9-2e2271d1dcf4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import regex as re\n",
    "import pylab as plt\n",
    "import pandas as pd\n",
    "\n",
    "from glob import glob\n",
    "from IPython import display\n",
    "from pathlib import Path\n",
    "from PIL import Image\n",
    "\n",
    "from tqdm.auto import tqdm"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6758c2c0-187e-4105-847e-4a0d79d2952e",
   "metadata": {},
   "source": [
    "## Settings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6f9bdd68-ae9f-4cd3-90af-bbf9a400f1c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download centerfolds first, e.g. from:\n",
    "# https://www.1377x.to/torrent/2758469/Playboy-Centerfolds-Ultra-High-Quality-The-Full-1953-2015-Year/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6f9bdd68-ae9f-4cd3-90af-bbf9a400f1c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir = 'data/centerfolds'"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ad567f9-f4d7-4006-818a-acb169c76028",
   "metadata": {},
   "source": [
    "## Extract metadata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0914be0a-d041-43d4-bc0f-dacecc1a7b6d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# img_files = sorted(glob(f'{data_dir}/**/*.jpg'))\n",
    "img_files = sorted(glob(f'{data_dir}/*197?*/*.jpg'))\n",
    "# img_files = sorted(glob(f'{data_dir}/*198?*/*.jpg'))\n",
    "albums = list(set([\n",
    "    Path(c).parent.stem\n",
    "    for c in img_files\n",
    "]))\n",
    "print(f\"Found {len(img_files)} number of images. This relates to {len(albums)} albums.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1f598997-725a-494a-8eee-23a41993fae2",
   "metadata": {},
   "outputs": [],
   "source": [
    "images = []\n",
    "widths = []\n",
    "heights = []\n",
    "for img_file in tqdm(img_files):\n",
    "    with Image.open(img_file) as im:\n",
    "        width, height = im.size\n",
    "        widths.append(width)\n",
    "        heights.append(height)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1ad6f340-e69f-4886-a9d9-f0bce3144350",
   "metadata": {},
   "outputs": [],
   "source": [
    "regex_year = re.compile(r'([0-9]{4})')\n",
    "regex_month = re.compile(r'([0-9]{2})_')\n",
    "data = {\n",
    "    'file_name': [Path(f).stem for f in img_files],\n",
    "    'file_path': [Path(f) for f in img_files],\n",
    "    'width': widths,\n",
    "    'height': heights,\n",
    "    # 'album': [Path(f).parent.stem for f in img_files],\n",
    "    'year': [re.findall(regex_year, Path(f).parent.stem)[0] for f in img_files],\n",
    "    'month': [re.findall(regex_month, Path(f).stem)[0] for f in img_files],\n",
    "}\n",
    "df = pd.DataFrame(columns=data.keys(), data=data)\n",
    "model_names = df['file_name'].str.extract(r'([A-Z][a-z]+)[ _]?([A-Z][a-z]+)(?:_CF)?$')\n",
    "df['model_name'] = model_names[0]\n",
    "df['model_surname'] = model_names[1]\n",
    "df = df.drop('file_name', axis=1)\n",
    "# Select only portrait formats\n",
    "len_df = len(df)\n",
    "df = df[df.width < df.height]\n",
    "print(f'Removed {len_df - len(df)} images in landscape format.')\n",
    "# Reindex\n",
    "df = df.reset_index()\n",
    "# Sort columns \n",
    "df = df[['year', 'month', 'model_name', 'model_surname', 'width', 'height', 'file_path']]\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4691bf09-06f7-4d3d-beff-e30c94c86e52",
   "metadata": {},
   "source": [
    "## Select images by resolution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d1dc4246-675b-477c-887f-89cb40eedf62",
   "metadata": {},
   "outputs": [],
   "source": [
    "import tabulate\n",
    "\n",
    "from IPython.display import display, HTML, Markdown"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4845fda5-adb7-4085-9672-35481c91d4e7",
   "metadata": {},
   "outputs": [],
   "source": [
    "min_w = df.width.min()\n",
    "min_h = df.height.min()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bcb79942-6fc1-4e3b-af4f-604023a86d2f",
   "metadata": {},
   "outputs": [],
   "source": [
    "bin_count = 50\n",
    "mode_w = df.width.mode()[0]\n",
    "mode_h = df.height.mode()[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "96f4e9ab-bc59-4e39-b447-4da419b68fee",
   "metadata": {},
   "outputs": [],
   "source": [
    "len_min_or = len(df[(df.height >= min_h) | (df.width >= min_w)])\n",
    "len_min_and = len(df[(df.height >= min_h) & (df.width >= min_w)])\n",
    "len_mode_or = len(df[(df.height >= mode_h) | (df.width >= mode_w)])\n",
    "len_mode_and = len(df[(df.height >= mode_h) & (df.width >= mode_w)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e89c9944-7bc6-461b-addc-56708f1dd521",
   "metadata": {},
   "outputs": [],
   "source": [
    "table_headers = [\"Selector\", \"Gate\", \"Width\", \"Height\", \"Selected\", \"Total\"]\n",
    "table_data = [\n",
    "    [\"Min\", \"OR\", f\"{min_w}\", f\"{min_h}\", f\"{len_min_or}\", f\"{len(df)}\"],\n",
    "    [\"Min\", \"AND\", f\"{min_w}\", f\"{min_h}\", f\"{len_min_and}\", f\"{len(df)}\"],\n",
    "    [\"Mode\", \"OR\", f\"{mode_w}\", f\"{mode_h}\", f\"{len_mode_or}\", f\"{len(df)}\"],\n",
    "    [\"Mode\", \"AND\", f\"{mode_w}\", f\"{mode_h}\", f\"{len_mode_and}\", f\"{len(df)}\"],\n",
    "]\n",
    "table = tabulate.tabulate(table_data, headers=table_headers, tablefmt='html')\n",
    "table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a648d767-f109-44e6-8d05-3258b13bee12",
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, axs = plt.subplots(1, 2)\n",
    "axs[0].hist(df.width, bins=bin_count)\n",
    "axs[0].set_title(\"Width\")\n",
    "axs[0].axvline(mode_w, label=f\"Mode @ {mode_w}\", color='C1')\n",
    "axs[0].axvline(min_w, label=f\"Min @ {min_w}\", color='C2')\n",
    "axs[0].legend()\n",
    "axs[1].hist(df.height, bins=bin_count)\n",
    "axs[1].set_title(\"Height\")\n",
    "axs[1].axvline(mode_h, label=f\"Mode @ {mode_h}\", color='C1')\n",
    "axs[1].axvline(min_h, label=f\"Min @ {min_h}\", color='C2')\n",
    "axs[1].legend()\n",
    "plt.suptitle(\"Occurences of image sizes per axis\")\n",
    "\n",
    "display(HTML(table))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e9a62eb-4a25-4f88-a5cd-68b707e79e73",
   "metadata": {},
   "source": [
    "### Get common image ratio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a0fdbd19-0889-42b7-9f68-1965ad1c5ca9",
   "metadata": {},
   "outputs": [],
   "source": [
    "from fractions import Fraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "24ebb133-2c46-48a2-a983-0b842153242c",
   "metadata": {},
   "outputs": [],
   "source": [
    "mode_ratio = Fraction(mode_w, mode_h)\n",
    "ratio_w, ratio_h = mode_ratio.numerator, mode_ratio.denominator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e9902fbe-1064-485b-89e4-08bf5dba7ab3",
   "metadata": {},
   "outputs": [],
   "source": [
    "display(Markdown(f\"(ℹ) Images are commonly shot in {ratio_w}:{ratio_h} format.\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e298daa-59f9-4c8f-b25a-866bc01328dd",
   "metadata": {},
   "source": [
    "## Pre-process images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f7219bbc-e3a3-4d2c-a146-252e03be2782",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_image_size(row, mode_w, mode_h, ratio_w, ratio_h):\n",
    "    ratio = ratio_w / ratio_h\n",
    "    fn = Path(row.file_path).stem\n",
    "    im = Image.open(row.file_path)\n",
    "    w, h = im.size\n",
    "    r = w / h\n",
    "    \n",
    "    # Rescale if size does not fit\n",
    "    if w != mode_w or h != mode_h:\n",
    "        assert ratio != r\n",
    "        # Select side depending on ratio\n",
    "        if ratio > r:\n",
    "            # Image taller, scale width\n",
    "            new_h = int(mode_w / r)\n",
    "            im = im.resize((mode_w, new_h))\n",
    "            w, h = im.size\n",
    "            # Crop height\n",
    "            crop_h = h - mode_h\n",
    "            crop_top = int(crop_h // 2)\n",
    "            crop_bottom = crop_top if crop_h % 2 == 0 else crop_top + 1\n",
    "            im = im.crop((0, crop_top, w, h - crop_bottom))\n",
    "        elif ratio < r:\n",
    "            # Image wider, scale height\n",
    "            new_w = int(mode_h * r)\n",
    "            im = im.resize((new_w, mode_h))\n",
    "            w, h = im.size\n",
    "            # Crop width\n",
    "            crop_w = w - mode_w\n",
    "            crop_left = int(crop_w // 2)\n",
    "            crop_right = crop_left if crop_w % 2 == 0 else crop_left + 1\n",
    "            im = im.crop((crop_left, 0, w - crop_right, h))\n",
    "        else:\n",
    "            print(f\"Rescaling using same ratio... yet untested\") \n",
    "            im = im.resize((mode_w, mode_h))\n",
    "    return im"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0d113435-bfcf-4c80-90f0-0f5c1876008f",
   "metadata": {},
   "outputs": [],
   "source": [
    "ims = {}\n",
    "for idx, row in tqdm(df.iterrows(), total=len(df)):\n",
    "    im = load_image_size(row, mode_w, mode_h, ratio_w, ratio_h)\n",
    "    ims[row.file_path] = im"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4361b5a2-4608-418a-8b2c-ca057bf3c830",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "daab86a1-b8ca-4c62-9bbe-5e707853a3dd",
   "metadata": {},
   "outputs": [],
   "source": [
    "mode_w, min_w"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b060b332-15dd-424e-bc75-8bb226215356",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4a6c9a16-726e-4cae-b8b4-63c7e6b0b393",
   "metadata": {},
   "outputs": [],
   "source": [
    "new_width, new_height = (w / 3, h / 2)\n",
    "\n",
    "left = (width - new_width) / 2\n",
    "top = (height - new_height) / 2\n",
    "right = (width + new_width) / 2\n",
    "bottom = (height + new_height) / 2\n",
    "\n",
    "# Crop the center of the image\n",
    "im = im.crop((left, top, right, bottom))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d283fee3-06ca-4b37-91e5-f80be2f5b023",
   "metadata": {},
   "outputs": [],
   "source": [
    "im.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "34462965-948d-4acf-ac98-da4f67502b6e",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}