examples/1c_shape-dataset-presentation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Shape dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook aims at presenting the Shapes dataset, a random set of images that depict simple geometrical shapes (*i.e.* squares, triangles, circles). This dataset is inspired from [Mask_RCNN](https://github.com/matterport/Mask_RCNN/tree/master/samples/shapes) repo, where a similar Shape dataset is built for illustration purpose."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Introduction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some modules are imported to do the job, as usual:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "import pandas as pd\n",
    "import json\n",
    "from PIL import Image\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: DEEPOSL_CONFIG=../config.ini\n"
     ]
    }
   ],
   "source": [
    "env DEEPOSL_CONFIG=../config.ini"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from deeposlandia import utils\n",
    "from deeposlandia.datasets import shapes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATAPATH = \"../data\"\n",
    "DATASET = \"shapes\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Metadata presentation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataset contains some images, splitted in training, validation and testing subsets on the same way as [Mapillary](./1a_mapillary-dataset-presentation.ipynb) dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The images here are `64*64` sized; they are composed of a square (or not), a circle (or not) and a triangle (or not). One can have empty images as well as images with three shapes. Then the shapes may overlap. The background as well as the shapes are randomly coloured. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training images: 100\n",
      "Number of validation images: 10\n",
      "Number of testing images: 10\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of training images: {}\".format(len(os.listdir(os.path.join(DATAPATH, DATASET, \"preprocessed\", \"64\", \"training\", \"images\")))))\n",
    "print(\"Number of validation images: {}\".format(len(os.listdir(os.path.join(DATAPATH, DATASET, \"preprocessed\", \"64\", \"validation\", \"images\")))))\n",
    "print(\"Number of testing images: {}\".format(len(os.listdir(os.path.join(DATAPATH, DATASET, \"preprocessed\", \"64\", \"testing\", \"images\")))))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Some examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show_example_image(image_filename):\n",
    "    im = Image.open(os.path.join(DATAPATH, DATASET, \"preprocessed\", \"64\", \"training\", \"images\", image_filename))\n",
    "    lab = Image.open(os.path.join(DATAPATH, DATASET, \"preprocessed\", \"64\", \"training\", \"labels\", image_filename))\n",
    "    f, ax = plt.subplots(1, 2, figsize=(12, 6))\n",
    "    ax[0].imshow(im)\n",
    "    ax[1].imshow(lab)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_example_image(\"shape_00002.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsMAAAFTCAYAAADcAgGcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAFWZJREFUeJzt3W+MZXd5H/DvEy+EdOMwtqHLyja1Kyz+vKgNWrkgUNTgEhESBV4gBImqpbJkqaQVaSMlppWqRmoleBOC1IJkBeKtRAOUhNqyUhLXcRRVqgxrbBL/gdhxDV5r7SWGSahfpDE8fTHHzcTs7tyZuTN3zv4+H2k095x779zv7lw//u6Zc+ZX3R0AABjRD606AAAArIoyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADD2lUZrqq3V9XXq+rRqrp5WaEAWD4zG+AH1U5XoKuqi5L8aZK3JTmV5MtJ3tfdD53zOZce7lx56Y5eD3biomdP7fi53zt8xRKTMHtPfDv97Wdr1TF2aicze21trY8ePbpPCSH5Zr654+e+Mq9cYhLm7vTp01lfX19oZh/axetcn+TR7n4sSarqM0nemeScgzVXXpof+u//chcvCdvzY1/6lR0/9y+u917lb3z/pz666gi7te2ZffTo0dx66637kw6SfCAf2PFzP56PLzEJc/f+979/4cfu5jSJy5M8sWn71LQPgIPHzAY4iz2/gK6qbqqqk1V1Ms88u9cvB8AubJ7Z6+vrq44DsOd2U4afTHLlpu0rpn1/S3ff0t3HuvtYLju8i5cDYBe2PbPX1tb2LRzAquzmnOEvJ7mmqq7OxkB9b5KfW0oq2IGX7uL84EW+3l9c/5Glfn3YZ2Y2B8puzg9e5Os5h5hF7bgMd/dzVfXPk/xekouSfKq7H1xaMgCWxswGOLvdHBlOd/9ukt9dUhYA9pCZDfCDrEAHAMCwlGEAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYu/rVarBKy15kYyevZyEOgMUse5GNnbyehTg4G0eGAQAYljIMAMCwlGEAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYyjAAAMNShgEAGJYV6JiF/V5tblFWpQP4Qfu92tyirErH2TgyDADAsJRhAACGpQwDADAsZRgAgGG5gI4D56BeLAfADzqoF8vBohwZBgBgWMowAADDUoYBABiWc4ZhySzEATAfFuLAkWEAAIalDAMAMCxlGACAYSnDAAAMywV0rJQFNgDmwwIbXIgcGQYAYFjKMAAAw1KGAQAYljIMAMCwXEAH++CFFwpakQ7g4HrhhYJWpLuwOTIMAMCwlGEAAIalDAMAMKwty3BVfaqqzlTVA5v2XVpVd1bVI9PnS/Y2JgCLMrcBFrfIBXS3JvmPSf7zpn03J7mruz9cVTdP25YSY0tWnIN9cWvMbZbAinOMYMsjw939R0m+/YLd70xyYrp9Ism7lpwLgB0ytwEWt9Nzho909+np9lNJjpzrgVV1U1WdrKqTeebZHb4cALu00NzePLPX19f3Lx3Aiuz6Arru7iR9nvtv6e5j3X0slx3e7csBsEvnm9ubZ/ba2to+JwPYfztddOPpqjra3aer6miSM8sMxYXB+cHndra/GwtxsMfMbc7L+cHndra/GwtxXDh2emT49iTHp9vHk9y2nDgA7BFzG+AsFvnVar+V5H8leXVVnaqqG5N8OMnbquqRJP942gbgADC3ARa35WkS3f2+c9x1w5KzALAE5jbA4qxABwDAsHZ6AR2wZC6qA5gPF9VdOBwZBgBgWMowAADDUoYBABiWMgwAwLBcQMdSWG0OYD6sNgd/w5FhAACGpQwDADAsZRgAgGEpwwAADMsFdHCAWZUOYD6sSjdPjgwDADAsZRgAgGEpwwAADMs5w2ybBTYA5sMCG3B+jgwDADAsZRgAgGEpwwAADEsZBgBgWC6gY0sumDtYXvj9sAgHsJkL5g6WF34/LMJx8DgyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCsQMffYrW5+Tnb98yqdDAGq83Nz9m+Z1alWy1HhgEAGJYyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWFuW4aq6sqrurqqHqurBqvrgtP/Sqrqzqh6ZPl+y93EBOB8zG2B7Fll047kkv9TdX6mqi5PcW1V3Jnl/kru6+8NVdXOSm5NYsWFGLLBx4bIQx9DM7AuUBTYuXBbiWK0tjwx39+nu/sp0+7tJHk5yeZJ3JjkxPexEknftVUgAFmNmA2zPts4Zrqqrkrw+yT1JjnT36emup5IcWWoyAHbFzAbY2sJluKp+NMlvJ/nF7v7Lzfd1dyfpczzvpqo6WVUn88yzuwoLwGKWMbPX19f3ISnAai1UhqvqRdkYqp/u7t+Zdj9dVUen+48mOXO253b3Ld19rLuP5bLDy8gMwHksa2avra3tT2CAFdryArqqqiSfTPJwd//aprtuT3I8yYenz7dt9bUu/t7Lc/13/9kOo3Iud1/8iR09zwVVcOFZ5sz+5jeTD7hma+k+vsProlxQBXtjkd8m8eYk/yTJn1TV/dO+f52Ngfq5qroxyTeSvGdvIgKwDWY2wDZsWYa7+38mqXPcfcNy4wCwG2Y2wPZYgQ4AgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAYljIMAMCwlGEAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYyjAAAMNShgEAGJYyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAYljIMAMCwlGEAAIa1ZRmuqpdU1Zeq6qtV9WBV/eq0/+qquqeqHq2qz1bVi/c+LgDnY2YDbM8iR4b/Kslbu/vaJNcleXtVvTHJR5J8tLtfleQ7SW7cu5gALMjMBtiGLctwb/g/0+aLpo9O8tYkn5/2n0jyrj1JCMDCzGyA7VnonOGquqiq7k9yJsmdSf4syXp3Pzc95FSSy8/x3Juq6mRVnfzr7/z5MjIDcB7LmtnPPbe+P4EBVmihMtzd3+vu65JckeT6JK9Z9AW6+5buPtbdx150yct2GBOARS1rZh86tLZnGQEOim39NonuXk9yd5I3JVmrqkPTXVckeXLJ2QDYBTMbYGuL/DaJl1fV2nT7R5K8LcnD2Riw754edjzJbXsVEoDFmNkA23No64fkaJITVXVRNsrz57r7jqp6KMlnqurfJ7kvySf3MCcAizGzAbZhyzLc3X+c5PVn2f9YNs5FA+CAMLMBtscKdAAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAYljIMAMCwlGEAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYyjAAAMNShgEAGJYyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAYljIMAMCwlGEAAIa1cBmuqouq6r6qumPavrqq7qmqR6vqs1X14r2LCcB2mNkAi9nOkeEPJnl40/ZHkny0u1+V5DtJblxmMAB2xcwGWMChRR5UVVck+ekk/yHJv6qqSvLWJD83PeREkn+X5BPn+zrfvehbufvi8z4EgF1a1sx+5SuTj398D4MCHACLHhn+9SS/nOT70/ZlSda7+7lp+1SSy5ecDYCdMbMBFrRlGa6qn0lyprvv3ckLVNVNVXWyqk7mmWd38iUAWNAyZ/b6+vqS0wEcPIucJvHmJD9bVe9I8pIkP5bkY0nWqurQdKThiiRPnu3J3X1LkluSpK69speSGoBzWdrMfu1rX2tmAxe8LY8Md/eHuvuK7r4qyXuT/EF3/3ySu5O8e3rY8SS37VlKABZiZgNsz25+z/CvZOPCjEezcT7aJ5cTCYA9YGYDnMVCv03ied39h0n+cLr9WJLrlx8JgGUwswG2ZgU6AACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAYljIMAMCwlGEAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYyjAAAMNShgEAGJYyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAY1qFFHlRVjyf5bpLvJXmuu49V1aVJPpvkqiSPJ3lPd39nb2ICsCgzG2Bx2zky/BPdfV13H5u2b05yV3dfk+SuaRuAg8HMBljAbk6TeGeSE9PtE0netfs4AOwRMxvgLBYtw53k96vq3qq6adp3pLtPT7efSnJk6ekA2AkzG2BBC50znOQt3f1kVf3dJHdW1dc239ndXVV9tidOg3hjGF9+yW6yArCYpczsV7ziFXufFGDFFjoy3N1PTp/PJPlCkuuTPF1VR5Nk+nzmHM+9pbuPdfexXHZ4OakBOKdlzey1tbX9igywMluW4ao6XFUXP387yU8meSDJ7UmOTw87nuS2vQoJwGLMbIDtWeQ0iSNJvlBVzz/+v3T3F6vqy0k+V1U3JvlGkvfsXUwAFmRmA2zDlmW4ux9Lcu1Z9j+T5Ia9CAXAzpjZANtjBToAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYyjAAAMNShgEAGJYyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw1KGAQAYljIMAMCwlGEAAIalDAMAMCxlGACAYSnDAAAMSxkGAGBYyjAAAMNShgEAGJYyDADAsJRhAACGpQwDADAsZRgAgGEpwwAADEsZBgBgWMowAADDWqgMV9VaVX2+qr5WVQ9X1Zuq6tKqurOqHpk+X7LXYQHYmpkNsLhFjwx/LMkXu/s1Sa5N8nCSm5Pc1d3XJLlr2gZg9cxsgAVtWYar6qVJfjzJJ5Oku/9vd68neWeSE9PDTiR5116FBGAxZjbA9ixyZPjqJN9K8ptVdV9V/UZVHU5ypLtPT495KsmRvQoJwMLMbIBtWKQMH0ryhiSf6O7XJ3k2L/jxWnd3kj7bk6vqpqo6WVUn88yzu80LwPktbWavr6/veViAVVukDJ9Kcqq775m2P5+NQft0VR1NkunzmbM9ubtv6e5j3X0slx1eRmYAzm1pM3ttbW1fAgOs0pZluLufSvJEVb162nVDkoeS3J7k+LTveJLb9iQhAAszswG259CCj/sXST5dVS9O8liSf5qNIv25qroxyTeSvGdvIgKwTWY2wIIWKsPdfX+SY2e564blxgFgt8xsgMVZgQ4AgGEpwwAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBhKcMAAAxLGQYAYFjKMAAAw6ru3r8Xq/pWkm8keVmSP9+3F16+Oeefc/Zk3vnnnD2R/+9198uXFWYOzOwDY87555w9mXf+OWdP9nFm72sZ/v8vWnWyu4/t+wsvyZzzzzl7Mu/8c86eyD+yuf/dyb86c86ezDv/nLMn+5vfaRIAAAxLGQYAYFirKsO3rOh1l2XO+eecPZl3/jlnT+Qf2dz/7uRfnTlnT+adf87Zk33Mv5JzhgEA4CBwmgQAAMPa9zJcVW+vqq9X1aNVdfN+v/52VdWnqupMVT2wad+lVXVnVT0yfb5klRnPpaqurKq7q+qhqnqwqj447T/w+avqJVX1par66pT9V6f9V1fVPdP757NV9eJVZz2fqrqoqu6rqjum7Vnkr6rHq+pPqur+qjo57Tvw75vnVdVaVX2+qr5WVQ9X1ZvmlP8gMbP3z5xndnJhzO25zuxk3nN71TN7X8twVV2U5D8l+akkr0vyvqp63X5m2IFbk7z9BftuTnJXd1+T5K5p+yB6LskvdffrkrwxyS9Mf99zyP9XSd7a3dcmuS7J26vqjUk+kuSj3f2qJN9JcuMKMy7ig0ke3rQ9p/w/0d3XbfrVNnN43zzvY0m+2N2vSXJtNr4Hc8p/IJjZ+27OMzu5MOb2nGd2Mt+5vdqZ3d379pHkTUl+b9P2h5J8aD8z7DD3VUke2LT99SRHp9tHk3x91RkX/HPcluRtc8uf5O8k+UqSf5iNX8B96Gzvp4P2keSK6T/gtya5I0nNJX+Sx5O87AX7ZvG+SfLSJP870zURc8t/kD7M7JX/OWY5s6ecs5vbc57ZU75Zzu2DMLP3+zSJy5M8sWn71LRvbo509+np9lNJjqwyzCKq6qokr09yT2aSf/px1f1JziS5M8mfJVnv7uemhxz098+vJ/nlJN+fti/LfPJ3kt+vqnur6qZp3yzeN0muTvKtJL85/bjzN6rqcOaT/yAxs1dkjjM7mf3cnvPMTuY7t1c+s11At0u98U+WA/0rOarqR5P8dpJf7O6/3HzfQc7f3d/r7uuy8a/165O8ZsWRFlZVP5PkTHffu+osO/SW7n5DNn48/gtV9eOb7zzI75skh5K8Icknuvv1SZ7NC368dsDzs4fm8L2f68xO5ju3L4CZncx3bq98Zu93GX4yyZWbtq+Y9s3N01V1NEmmz2dWnOecqupF2Riqn+7u35l2zyZ/knT3epK7s/EjqrWqOjTddZDfP29O8rNV9XiSz2Tjx24fy0zyd/eT0+czSb6Qjf+pzeV9cyrJqe6+Z9r+fDYG7VzyHyRm9j67EGZ2Msu5PeuZncx6bq98Zu93Gf5ykmumqzNfnOS9SW7f5wzLcHuS49Pt49k4r+vAqapK8skkD3f3r22668Dnr6qXV9XadPtHsnHe3MPZGK7vnh52ILMnSXd/qLuv6O6rsvE+/4Pu/vnMIH9VHa6qi5+/neQnkzyQGbxvkqS7n0ryRFW9etp1Q5KHMpP8B4yZvY/mPLOTec/tOc/sZN5z+0DM7BWcKP2OJH+ajfOI/s1+v/4O8v5WktNJ/job/3q5MRvnEd2V5JEk/yPJpavOeY7sb8nGjxX+OMn908c75pA/yT9Ict+U/YEk/3ba//eTfCnJo0n+a5IfXnXWBf4s/yjJHXPJP2X86vTx4PP/nc7hfbPpz3BdkpPT++e/JblkTvkP0oeZva/ZZzuzp/wXxNye28zelHO2c3vVM9sKdAAADMsFdAAADEsZBgBgWMowAADDUoYBABiWMgwAwLCUYQAAhqUMAwAwLGUYAIBh/T+1aZpZVVwUfAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_example_image(\"shape_00000.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_example_image(\"shape_00033.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_example_image(\"shape_00042.png\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dataset handling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "One may create a shape dataset easily in two steps. First one has to declare an instance of `ShapeDataset`, by passing an image size as a parameter."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "d = shapes.ShapeDataset(48)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "At this step, there is not any image in the dataset, however, the label glossary is seeable."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['square', 'circle', 'triangle', 'background']"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "[item['name'] for item in d.label_info]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Then, the dataset must be fillet out with some brand new images. One may specify the image quantity, as well as an output directory if one wants to write these images onto the disk:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "d.populate(output_dir=None, nb_images=5)\n",
    "d.get_nb_images()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataset may also be loaded starting from an existing dataset: by picking a `.json` configuration file stored in the `preprocessed` folder, one may recover some previously generated files."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2019-05-28 10:59:42,841 :: INFO :: __init__ :: load : The dataset has been loaded from ../data/shapes/preprocessed/64/training.json\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocess_folder = utils.prepare_preprocessed_folder(DATAPATH, DATASET, 64)\n",
    "d.load(preprocess_folder[\"training_config\"], 5)\n",
    "d.get_nb_images()"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}