{
 "cells": [
  {
   "cell_type": "code",
   "id": "9ca8370e29b99d9b",
   "metadata": {
    "collapsed": true,
    "ExecuteTime": {
     "end_time": "2024-04-17T07:40:35.575275Z",
     "start_time": "2024-04-17T07:40:35.397474Z"
    }
   },
   "source": [
    "import json\n",
    "import os\n",
    "import shutil\n",
    "\n",
    "import pandas as pd"
   ],
   "execution_count": 2,
   "outputs": []
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-04-17T07:59:30.461558Z",
     "start_time": "2024-04-17T07:59:30.458532Z"
    }
   },
   "cell_type": "code",
   "source": [
    "def load_dataset(input_annotations_file: str, input_img_dir: str, output_dir: str):\n",
    "    with open(input_annotations_file, \"r\") as f:\n",
    "        dataset = json.load(f)\n",
    "    df_annotation = pd.DataFrame(dataset['annotations']).drop(['segmentation', 'attribute_ids'], axis='columns')\n",
    "    df_images = pd.DataFrame(dataset['images'])\n",
    "    df_categories = pd.DataFrame(dataset['categories'])\n",
    "\n",
    "    df_dataset = pd.merge(df_images, df_annotation, how='inner', left_on='id', right_on='image_id')\n",
    "    df_dataset = pd.merge(df_dataset, df_categories, how='inner', left_on='category_id', right_on='id')\n",
    "    df_dataset['path_to_image'] = input_img_dir + df_dataset['file_name']\n",
    "    return df_dataset\n"
   ],
   "id": "initial_id",
   "execution_count": 48,
   "outputs": []
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-04-17T07:59:30.763506Z",
     "start_time": "2024-04-17T07:59:30.759581Z"
    }
   },
   "cell_type": "code",
   "source": [
    "def save_labels_one_image(df_one_image: pd.DataFrame, output_folder: str, subfolder: str):\n",
    "    path_to_image = df_one_image.iloc[0]['path_to_image']\n",
    "    image_id = df_one_image.iloc[0]['image_id']\n",
    "    image_width = df_one_image.iloc[0]['width']\n",
    "    image_height = df_one_image.iloc[0]['height']\n",
    "\n",
    "    def format_one_box(one_line) -> str:\n",
    "        label_index = str(one_line['category_id'])\n",
    "\n",
    "        [x_center, y_center, width, height] = one_line['bbox']\n",
    "        x_center_normalised, y_center_normalised = (x_center+width/2) / image_width, (y_center+height/2) / image_height\n",
    "        width_normalised, height_normalised = width / image_width, height / image_height\n",
    "        box = ' '.join(map(str, [x_center_normalised, y_center_normalised, width_normalised, height_normalised]))\n",
    "\n",
    "        line = f'{label_index} {box}'\n",
    "        return line\n",
    "\n",
    "    boxes_formatted = df_one_image.apply(format_one_box, axis='columns').tolist()\n",
    "\n",
    "    with open(os.path.join(output_folder, 'labels', subfolder, f'{image_id}.txt'), 'w') as f:\n",
    "        f.write('\\n'.join(boxes_formatted))\n",
    "    shutil.copy(path_to_image, os.path.join(output_folder, 'images', subfolder, f'{image_id}.jpg'))\n",
    "\n",
    "    return image_id\n"
   ],
   "id": "aa0ec50c0b8b00f3",
   "execution_count": 49,
   "outputs": []
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-04-17T07:59:33.282829Z",
     "start_time": "2024-04-17T07:59:32.281139Z"
    }
   },
   "cell_type": "code",
   "source": [
    "path_to_json_validation = \"./dataset/fashionpedia2020/instances_attributes_val2020.json\"\n",
    "path_to_image_folder_validation = './dataset/fashionpedia2020/val_test2020/test/'\n",
    "output_folder_validation = 'val2020'\n",
    "fashionpedia_dataset_validation = load_dataset(path_to_json_validation, path_to_image_folder_validation,\n",
    "                                               'dataset/fashionpedia-yolo')\n",
    "fashionpedia_dataset_validation.groupby(['image_id']).apply(save_labels_one_image, './dataset/fashionpedia-yolo',\n",
    "                                                            output_folder_validation)"
   ],
   "id": "31756215355a324e",
   "execution_count": 50,
   "outputs": []
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-04-17T08:00:11.718962Z",
     "start_time": "2024-04-17T07:59:33.326724Z"
    }
   },
   "cell_type": "code",
   "source": [
    "path_to_json_train = \"./dataset/fashionpedia2020/instances_attributes_train2020.json\"\n",
    "path_to_image_folder_train = './dataset/fashionpedia2020/train2020/train/'\n",
    "output_folder_train = 'train2020'\n",
    "fashionpedia_dataset_train = load_dataset(path_to_json_train, path_to_image_folder_train, 'dataset/fashionpedia-yolo')\n",
    "fashionpedia_dataset_train.groupby(['image_id']).apply(save_labels_one_image, './dataset/fashionpedia-yolo',\n",
    "                                                       output_folder_train)"
   ],
   "id": "ba95b509750ea10b",
   "execution_count": 51,
   "outputs": []
  },
  {
   "metadata": {},
   "cell_type": "code",
   "execution_count": null,
   "source": "",
   "id": "77dff2cb00f9b08c",
   "outputs": []
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-04-17T08:00:11.733525Z",
     "start_time": "2024-04-17T08:00:11.731305Z"
    }
   },
   "cell_type": "code",
   "source": [
    "from matplotlib import pyplot as plt\n",
    "\n",
    "COLORS = ['lightcoral', 'yellowgreen', 'darkturquoise', 'hotpink', 'mediumslateblue']\n",
    "\n",
    "\n",
    "def plot_results(pil_img, boxes, class_labels):\n",
    "    plt.figure(figsize=(16, 10))\n",
    "    plt.imshow(pil_img)\n",
    "    ax = plt.gca()\n",
    "    colors = COLORS * 100\n",
    "    for [xcenter, ycenter, xlength, ylength], c, label in zip(boxes, colors, class_labels):\n",
    "        xmin = (xcenter-(xlength/2)) * pil_img.width\n",
    "        xmax = (xcenter+(xlength/2)) * pil_img.width\n",
    "        ymin = (ycenter-(ylength/2)) * pil_img.height\n",
    "        ymax = (ycenter+(ylength/2)) * pil_img.height\n",
    "        \n",
    "        print(xmin, ymin)\n",
    "        ax.add_patch(plt.Rectangle((xmin, ymin), xmax - xmin, ymax - ymin,\n",
    "                                   fill=False, color=c, linewidth=2))\n",
    "        ax.text(xmin, ymin, str(label), fontsize=15,\n",
    "                bbox=dict(facecolor=c, alpha=0.5))\n",
    "    plt.axis('off')\n",
    "    plt.show()"
   ],
   "id": "a16ebfa1cf517c1a",
   "execution_count": 52,
   "outputs": []
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-04-17T08:00:11.921853Z",
     "start_time": "2024-04-17T08:00:11.733970Z"
    }
   },
   "cell_type": "code",
   "source": [
    "from PIL import Image\n",
    "\n",
    "pil_image = Image.open('dataset/fashionpedia-yolo/images/train2020/23.jpg')\n",
    "with open('dataset/fashionpedia-yolo/labels/train2020/23.txt' ) as f:\n",
    "    labels = f.readlines()\n",
    "\n",
    "class_labels  = []\n",
    "box_center_length = []\n",
    "for a_label in labels:\n",
    "    labels_split = a_label.split(' ')\n",
    "    class_labels.append(int(labels_split[0]))\n",
    "    box_center_length.append(list(map(float, labels_split[1:])))\n",
    "plot_results(pil_image,box_center_length, class_labels)"
   ],
   "id": "8c5dacbcecca7d42",
   "execution_count": 53,
   "outputs": []
  },
  {
   "metadata": {},
   "cell_type": "code",
   "execution_count": null,
   "source": "",
   "id": "d7ac9abea09107e1",
   "outputs": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}