build_dataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "'''This script demonstrates how to build the data set for rubiks\n",
    "'''\n",
    "import warnings\n",
    "import sys, os\n",
    "from glob import glob\n",
    "import numpy as np\n",
    "np.random.seed(0)\n",
    "import matplotlib.pyplot as plt\n",
    "from scipy.stats import norm\n",
    "from skimage.transform import resize\n",
    "import csv\n",
    "\n",
    "data_dir = 'asset'\n",
    "num_data = 19 # samples per viewpoint use for augmentation\n",
    "\n",
    "# These values are allowed to change\n",
    "num_tuples = int(10000) # number of actions which will be created\n",
    "target_size = (30, 40)  # target image size\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define \n",
    "\n",
    "sample_folders =  ['bag_front_state_blue',\n",
    "'bag_front_state_green',\n",
    "'bag_front_state_orange',\n",
    "'bag_front_state_red',\n",
    "'bag_front_state_white',\n",
    "'bag_front_state_yellow',\n",
    "'bag_left_state_blue',\n",
    "'bag_left_state_green',\n",
    "'bag_left_state_orange',\n",
    "'bag_left_state_red',\n",
    "'bag_left_state_white',\n",
    "'bag_left_state_yellow',\n",
    "'bag_right_state_blue',\n",
    "'bag_right_state_green',\n",
    "'bag_right_state_orange',\n",
    "'bag_right_state_red',\n",
    "'bag_right_state_white',\n",
    "'bag_right_state_yellow']\n",
    "\n",
    "def sample_cube_state(num_samples):\n",
    "    ''' Sampling of the different cube states '''\n",
    "    cube_colors = ['red', 'white', 'green', 'orange', 'yellow', 'blue']\n",
    "    cube_opposit_sides = [['blue', 'green'], ['white','yellow'], ['orange','red'], ['green', 'blue'], ['yellow', 'white'], ['red', 'orange']]\n",
    "    cube_states = {'left': [], 'front': [], 'right': []}\n",
    "    # sample left color\n",
    "    lookup_cube_colors = np.random.randint(0, high=len(cube_colors), size=num_samples)\n",
    "    for idx in lookup_cube_colors:\n",
    "        # Add the left view color\n",
    "        cube_states['left'].append(cube_colors[idx])\n",
    "        # Get the color of the opposit side\n",
    "        cube_opposit_side_color = [color[1] for color in cube_opposit_sides if color[0] == cube_colors[idx]]\n",
    "        cube_states['right'].append(cube_opposit_side_color[0])\n",
    "        # Sample the front view from the remaining colors\n",
    "        cube_colors_wo_left_right = cube_colors.copy()\n",
    "        cube_colors_wo_left_right.remove(cube_states['left'][-1])\n",
    "        cube_colors_wo_left_right.remove(cube_states['right'][-1])\n",
    "        cube_states['front'].append(cube_colors_wo_left_right[np.random.randint(0, high=4, size=1)[0]])\n",
    "    return cube_states\n",
    "\n",
    "def get_iterator(batch_size, data_sub_dir = \"\"):\n",
    "    ''' Data set iterator '''\n",
    "    from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "    train_datagen = ImageDataGenerator(\n",
    "        rescale=1./255,\n",
    "        shear_range=0,\n",
    "        zoom_range=0,\n",
    "        horizontal_flip=False,\n",
    "        width_shift_range=0.0,  # randomly shift images horizontally (fraction of total width)\n",
    "        height_shift_range=0.0)  # randomly shift images vertically (fraction of total height))\n",
    "\n",
    "    train_generator = train_datagen.flow_from_directory(data_dir + \"/\" + data_sub_dir, interpolation='nearest',\n",
    "            color_mode='rgb', shuffle=False, seed=None,\n",
    "            target_size=target_size,\n",
    "            batch_size=batch_size,\n",
    "            #save_to_dir='img_0_augmented',\n",
    "            class_mode=None)\n",
    "    return train_generator\n",
    "\n",
    "   \n",
    "with open(data_dir + \"/actions.csv\", \"r\") as f:\n",
    "    reader = csv.reader(f, delimiter=',')\n",
    "    actions = [[x[0], x[1], float(x[2]), float(x[3]), float(x[4])] for x in list(reader)[1:]]\n",
    "    f.close()\n",
    "    \n",
    "with open(data_dir + \"/colors.csv\", \"r\") as f:\n",
    "    reader = csv.reader(f, delimiter=',')\n",
    "    colors = list(reader)[1:]\n",
    "    f.close()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 27 images belonging to 1 classes.\n",
      "Found 25 images belonging to 1 classes.\n",
      "Found 29 images belonging to 1 classes.\n",
      "Found 26 images belonging to 1 classes.\n",
      "Found 30 images belonging to 1 classes.\n",
      "Found 27 images belonging to 1 classes.\n",
      "Found 24 images belonging to 1 classes.\n",
      "Found 20 images belonging to 1 classes.\n",
      "Found 19 images belonging to 1 classes.\n",
      "Found 21 images belonging to 1 classes.\n",
      "Found 27 images belonging to 1 classes.\n",
      "Found 25 images belonging to 1 classes.\n",
      "Found 27 images belonging to 1 classes.\n",
      "Found 24 images belonging to 1 classes.\n",
      "Found 27 images belonging to 1 classes.\n",
      "Found 33 images belonging to 1 classes.\n",
      "Found 49 images belonging to 1 classes.\n",
      "Found 40 images belonging to 1 classes.\n"
     ]
    }
   ],
   "source": [
    "# Read the poses and observations\n",
    "X_train_set = {}\n",
    "for sample_folder in sample_folders:\n",
    "    X_train_set[sample_folder] = get_iterator(num_data, sample_folder).next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Show some scaled image\n",
    "plt.imshow(X_train_set['bag_front_state_blue'][0])\n",
    "plt.title('bag_front_state_blue')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Action Vp 1 -> Vp 2:  [0.5 1.  0. ]\n"
     ]
    }
   ],
   "source": [
    "# Create the action/perception tuples\n",
    "action_train = np.zeros((num_tuples, 3))\n",
    "v1_train = np.zeros((num_tuples, target_size[0], target_size[1], 3))\n",
    "v2_train = np.zeros((num_tuples, target_size[0], target_size[1], 3))\n",
    "cube_states = sample_cube_state(num_samples = num_tuples)\n",
    "\n",
    "lookup = np.random.randint(0, high=len(actions), size=num_tuples)\n",
    "lookup_cube_state = np.random.randint(0, high=num_data, size=num_tuples)\n",
    "\n",
    "for idx_lookup, idx in zip(lookup, range(num_tuples)):\n",
    "    # Get the action for the random sample\n",
    "    v1_text = actions[idx_lookup][0]\n",
    "    v2_text = actions[idx_lookup][1]\n",
    "    action_train[idx] = np.asarray(actions[idx_lookup][2:])\n",
    "    # Get the views for the action sample and a given cube state\n",
    "    random_vp_sample = np.random.randint(0, high=num_data, size=1)\n",
    "    v1_train[idx,:] = X_train_set['bag_' + v1_text + '_state_' + cube_states[v1_text][idx]][random_vp_sample[0]]\n",
    "    random_vp_sample = np.random.randint(0, high=num_data, size=1)\n",
    "    v2_train[idx,:] = X_train_set['bag_' + v2_text + '_state_' + cube_states[v2_text][idx]][random_vp_sample[0]]\n",
    "\n",
    "# Normalize actions\n",
    "action_train = (action_train + 1.) / 2.\n",
    "\n",
    "# Show a tuple\n",
    "idx = 0\n",
    "f, ax = plt.subplots(1,2,figsize=(7.5,7.5), dpi=96)\n",
    "ax[0].imshow(v1_train[idx])\n",
    "ax[1].imshow(v2_train[idx])\n",
    "ax[0].set_title('Vp 1: ' + actions[lookup[idx]][0])\n",
    "ax[1].set_title('Vp 2: ' + actions[lookup[idx]][1])\n",
    "plt.show()\n",
    "print(\"Action Vp 1 -> Vp 2: \", action_train[idx])"
   ]
  }
 ],
 "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": "3.5.2"
  },
  "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()) "
    }
   },
   "position": {
    "height": "693px",
    "left": "1442.97px",
    "right": "20px",
    "top": "172.969px",
    "width": "359px"
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}