{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_obj = pd.read_parquet('object_annotations.parquet')\n",
    "df_obj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a new column bbox with a list of the bounding box coordinates\n",
    "df_obj['bbox'] = df_obj.apply(lambda row: [row['col_x'], row['row_y'], row['width'], row['height']], axis=1)\n",
    "df_obj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_obj = df_obj.drop(columns=['col_x', 'row_y', 'width', 'height', 'caption', 'source'])\n",
    "df_obj "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# group by filenmae and aggregate the bbox, confidence, label columns into a list\n",
    "df_obj = df_obj.groupby('filename').agg({'bbox': list, 'confidence': list, 'label': list}).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_obj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a new column 'labels' with a dict of bbox, confidence, label\n",
    "df_obj['labels'] = df_obj.apply(lambda row: [{'bbox': bbox, 'confidence': confidence, 'label': label} for bbox, confidence, label in zip(row['bbox'], row['confidence'], row['label'])], axis=1)\n",
    "df_obj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# all columns except filename and labels\n",
    "df_obj = df_obj[['filename', 'labels']]\n",
    "df_obj"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plot the first row 'filename' and 'labels' column using matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as patches\n",
    "import numpy as np\n",
    "from PIL import Image\n",
    "\n",
    "row = df_obj.iloc[0]\n",
    "filename = row['filename']\n",
    "labels = row['labels']\n",
    "\n",
    "labels\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load the image\n",
    "img = Image.open(f'{filename}')\n",
    "img = np.array(img)\n",
    "\n",
    "# create figure and axes\n",
    "fig, ax = plt.subplots(1)\n",
    "\n",
    "# display the image\n",
    "ax.imshow(img)\n",
    "\n",
    "# plot bbox on the image along with labels and confidence\n",
    "for label in labels:\n",
    "    bbox = label['bbox']\n",
    "    rect = patches.Rectangle((bbox[0], bbox[1]), bbox[2], bbox[3], linewidth=1, edgecolor='r', facecolor='none')\n",
    "    ax.add_patch(rect)\n",
    "    ax.text(bbox[0], bbox[1], f\"{label['label']} {label['confidence']:.2f}\", color='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_parquet('image_annotations.parquet')\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop all columns except filename, caption, label, class id\n",
    "df = df[['filename', 'caption', 'label']]\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# merge with df_obj on filename\n",
    "df = df.merge(df_obj, on='filename')\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# rename label to image_label, class_id to image_class_id\n",
    "df.rename(columns={'label': 'image_labels', 'labels': 'objects'}, inplace=True)\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_parquet('annotations.parquet', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}