openimagesv5.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Open Image v5 Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.vision import *\n",
    "from fastprogress import progress_bar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!wget https://open-images-dataset.s3.amazonaws.com/tar/validation.tar.gz -P {Config.data_path()}\n",
    "!tar -xvf {Config.data_path()}/validation.tar.gz\n",
    "!mv {Config.data_path()}/validation  {Config.data_path()}/openimagesv5\n",
    "!wget https://storage.googleapis.com/openimages/v5/validation-annotations-bbox.csv -P {Config.data_path()}/openimagesv5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = Config.data_path() / 'openimagesv5'\n",
    "labels = pd.read_csv(path/'validation-annotations-bbox.csv')\n",
    "labels['ImageID'] = labels['ImageID']+'.jpg'  \n",
    "classes = pd.read_csv(path/'classes.csv', names=['ID', 'Name'])\n",
    "classes.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Get the real names of the classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in classes['ID'].unique():\n",
    "    labels.loc[labels['LabelName']==i, 'LabelName'] = classes[classes['ID'] == i]['Name'].iloc[0]\n",
    "labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "fastai requires the x, y points to be scaled to the image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "files = [path/x for x in labels['ImageID'].unique().tolist()]\n",
    "open_image(files[0]).px.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "w,h=1024,447"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels['XMin'] *= w\n",
    "labels['XMax'] *= w\n",
    "labels['YMin'] *= h\n",
    "labels['YMax'] *= h"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "ls = []\n",
    "\n",
    "for i in progress_bar(files):\n",
    "    id = Path(i).stem + '.jpg'\n",
    "    boxes = labels[labels['ImageID']==id]\n",
    "    # top, left, bottom, right\n",
    "    bounds = boxes[['YMin', 'XMin', 'YMax', 'XMax']].values.tolist()\n",
    "    l =  boxes[['LabelName']].values.tolist()\n",
    "    l = [x[0] for x in l]\n",
    "    ls.append((bounds, l))\n",
    "\n",
    "ls = dict(zip(files, ls))\n",
    "\n",
    "def labeller(x):    \n",
    "    return ls[x]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "data = ObjectItemList(files)\n",
    "ldata = (data.split_by_rand_pct(0.2, seed=2).label_from_func(labeller))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = ldata.transform(size=224, tfm_y=True).databunch(bs=2, collate_fn=bb_pad_collate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data.show_batch(rows=3, figsize=(9,9))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Basic Example Usage\n",
    "\n",
    "```py\n",
    "coco = untar_data(URLs.COCO_TINY)\n",
    "images, lbl_bbox = get_annotations(coco/'train.json')\n",
    "img2bbox = dict(list(zip(images, lbl_bbox)))\n",
    "get_y_func = lambda o:img2bbox[o.name]\n",
    "\n",
    "data = (ObjectItemList.from_folder(coco)\n",
    "        #Where are the images? -> in coco and its subfolders\n",
    "        .split_by_rand_pct()                          \n",
    "        #How to split in train/valid? -> randomly with the default 20% in valid\n",
    "        .label_from_func(get_y_func)\n",
    "        #How to find the labels? -> use get_y_func on the file name of the data\n",
    "        .transform(get_transforms(), tfm_y=True)\n",
    "        #Data augmentation? -> Standard transforms; also transform the label images\n",
    "        .databunch(bs=16, collate_fn=bb_pad_collate))   \n",
    "        #Finally we convert to a DataBunch, use a batch size of 16,\n",
    "        # and we use bb_pad_collate to collate the data into a mini-batch\n",
    "```"
   ]
  }
 ],
 "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": "3.6.8"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}