{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "df45e4ec-e732-4cf4-8017-ce3753f4cd48",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "%cd .."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "67eda6f2-e2e5-495a-8795-25365d46c081",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import re\n",
    "from pathlib import Path\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from realfake.utils import list_files, write_jsonl"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c6e57858-c5f1-4ef9-a69b-31e402215564",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "np.random.seed(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5f8e03d2-039d-4aaf-b8bd-173d04b8d888",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "root = Path(\"/fsx/home-devforfu/data\")\n",
    "laion, movie1, movie2 = [list_files(root/subdir, [\"jpg\"]) for subdir in (\"laionimages\", \"shotcafe\", \"pack1\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1766851a-e1f6-46af-98a1-700aa53dd240",
   "metadata": {},
   "outputs": [],
   "source": [
    "len(movie2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dcb3b8ab-6012-4e78-958b-aef1bc743ef9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "n_test = 0.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "419062b0-e594-4644-87f2-180ba864587e",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "tst = set(train_test_split(np.arange(len(laion)), test_size=n_test)[1])\n",
    "metadata = [{\"path\": str(fn), \"label\": \"real\", \"valid\": i in tst} for i, fn in enumerate(laion)]\n",
    "tst = set(train_test_split(np.arange(len(movie1)), test_size=n_test)[1])\n",
    "metadata += [{\"path\": str(fn), \"label\": \"fake\", \"valid\": i in tst} for i, fn in enumerate(movie1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8f6ed004-8932-4ce9-b4cb-32d0c5c47da7",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "movie_to_frame = defaultdict(list)\n",
    "for fn in movie2:\n",
    "    movie_name = re.search(\"((?:[a-zA-Z]+|[0-9]+))\", fn.stem).group(1)\n",
    "    movie_to_frame[movie_name].append(fn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bb06edea-419c-4e3f-b9dd-f4869e05d3a9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trn_keys, tst_keys = train_test_split(list(set(movie_to_frame)), test_size=n_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6ac623e7-c9ae-4467-9fc7-f363adae6e57",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "metadata += [{\"path\": str(fn), \"label\": \"fake\", \"valid\": False} for key in trn_keys for fn in movie_to_frame[key]]\n",
    "metadata += [{\"path\": str(fn), \"label\": \"fake\", \"valid\": True} for key in tst_keys for fn in movie_to_frame[key]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a70b6b30-fdbb-49cb-b10f-34547c4072bb",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "df = pd.DataFrame(metadata)\n",
    "print(df.valid.value_counts())\n",
    "print(df.label.value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "737a3a2f-428d-490f-a8a9-def76f094ce0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "pos_weight = (1 - df.label.value_counts(normalize=True)).tolist()\n",
    "pos_weight"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "72e03454-cec8-4b58-ba3a-f1fe7d59ee8e",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "write_jsonl(\"metadata/movies_plus.jsonl\", metadata)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "41c061b1-7f8d-43d5-9b52-88c6144d0bda",
   "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": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}