Haralick Feature Model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.image as mpimg\n",
    "%matplotlib inline\n",
    "from PIL import Image\n",
    "from scipy.misc import imread\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "# sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))\n",
    "# from keras import backend as K\n",
    "# K.set_session(sess)\n",
    "\n",
    "import os\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning)\n",
    "# warnings.filterwarnings(\"ignore\", category=UserWarning)\n",
    "# warnings.filterwarnings(\"ignore\", category=FutureWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.append('./utils')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_path = \"./dataset/train/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from load_labels import LoadLabels\n",
    "\n",
    "train_labels, label_names = LoadLabels('./dataset/train.csv').load_labels()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from partition_data import PartitionData\n",
    "partitionData = PartitionData(train_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "partitions = partitionData.RepeatedMultilabelStratifiedKFold(3,1,420)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from model_parameter import ModelParameter\n",
    "parameter = ModelParameter(train_path, shuffle=True, batch_size=3, n_channels=3, n_epochs=5, row_scale_factor=2, col_scale_factor=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import mahotas as mt\n",
    "from image_preprocessor import ImagePreprocessor\n",
    "# from scipy.misc import imread\n",
    "# from skimage.transform import resize\n",
    "# from skimage.color import rgb2grey\n",
    "\n",
    "# class ExtractImageFeatures:\n",
    "#     def __init__(self, modelparameter):\n",
    "#         self.parameter = modelparameter\n",
    "#         self.basepath = self.parameter.basepath\n",
    "#         self.scaled_row_dim = self.parameter.scaled_row_dim\n",
    "#         self.scaled_col_dim = self.parameter.scaled_col_dim\n",
    "#         self.n_channels = self.parameter.n_channels\n",
    "        \n",
    "#     def resize(self, image):\n",
    "#         image = resize(image, (self.scaled_row_dim, self.scaled_col_dim))\n",
    "#         return image\n",
    "    \n",
    "#     def normalize(self, image):\n",
    "#         image /= 255 \n",
    "#         return image\n",
    "        \n",
    "#     def preprocess(self, image_id):\n",
    "#         image = self.load_image(image_id)\n",
    "#         image = self.resize(image)\n",
    "#         textures = mt.features.haralick(image.astype(int))\n",
    "#         ht_mean = textures.mean(axis=0)\n",
    "#         return ht_mean\n",
    "    \n",
    "#     def load_image(self, image_id):\n",
    "#         image = np.zeros(shape=(512,512,3))\n",
    "#         image[:,:,2] = imread(self.basepath + image_id + \"_blue\" + \".png\")\n",
    "#         image[:,:,1] = imread(self.basepath + image_id + \"_green\" + \".png\")\n",
    "#         image[:,:,0] = imread(self.basepath + image_id + \"_red\" + \".png\")\n",
    "#         return image\n",
    "\n",
    "class ExtractImageFeatures(ImagePreprocessor):\n",
    "    def __init__(self, modelparameter):\n",
    "        super().__init__(modelparameter)\n",
    "        \n",
    "    def preprocess(self, image_id):\n",
    "        image = self.load_image(image_id)\n",
    "        image = self.resize(image)\n",
    "#         image = self.normalize(image)\n",
    "        textures = mt.features.haralick(image.astype(int))\n",
    "        ht_mean = textures.mean(axis=0)\n",
    "        return ht_mean\n",
    "    \n",
    "preprocessor = ExtractImageFeatures(parameter)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 2.70222072e-02  1.79288236e+02  1.21474270e-01  1.03004597e+02\n",
      "  2.23841041e-01  1.45070397e+01  2.32730151e+02  5.25336929e+00\n",
      "  7.94173351e+00  2.58551633e-04  4.55324115e+00 -5.12316001e-02\n",
      "  5.03556360e-01]\n"
     ]
    }
   ],
   "source": [
    "print(preprocessor.preprocess(partitions[0]['train'][0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_targets_per_image(identifier, labels):\n",
    "    return labels.loc[labels.Id==identifier].drop([\"Id\", \"Target\"], axis=1).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# generator function to return images batchwise\n",
    "def generator(listIDs, labels, modelparameter, imagepreprocessor):\n",
    "    while True:\n",
    "        # Randomize the indices to make an array\n",
    "        indices_arr = np.random.permutation(listIDs.shape[0])\n",
    "        for batch in range(0, len(indices_arr), modelparameter.batch_size):\n",
    "            # slice out the current batch according to batch-size\n",
    "            current_batch = indices_arr[batch:(batch + modelparameter.batch_size)]\n",
    "\n",
    "            # initializing the arrays, x_train and y_train\n",
    "            x_train = np.empty((modelparameter.batch_size, 13), dtype=np.float32)\n",
    "            y_train = np.empty((modelparameter.batch_size, modelparameter.num_classes), dtype=int)\n",
    "\n",
    "            for (i,ind) in enumerate(current_batch):\n",
    "                # Appending them to existing batch\n",
    "                x_train[i,] = preprocessor.preprocess(listIDs[ind])\n",
    "                # Store class\n",
    "                y_train[i] = get_targets_per_image(listIDs[ind], labels) \n",
    "\n",
    "            yield (x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "genrtor = generator(partitions[0]['train'], train_labels, parameter, preprocessor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([[ 2.33969793e-01,  1.62238708e+03,  4.69740510e-01,\n",
       "          1.51661755e+03,  5.19838333e-01,  3.45970993e+01,\n",
       "          4.44408301e+03,  4.75413465e+00,  6.39355183e+00,\n",
       "          9.50294023e-04,  4.23160696e+00, -1.41580105e-01,\n",
       "          6.95594549e-01],\n",
       "        [ 3.64103377e-01,  1.93865979e+03,  3.02991211e-01,\n",
       "          1.34007764e+03,  6.29073918e-01,  3.10234032e+01,\n",
       "          3.42165063e+03,  3.95687437e+00,  5.59849024e+00,\n",
       "          1.44431018e-03,  3.62597919e+00, -1.79732561e-01,\n",
       "          7.78283536e-01],\n",
       "        [ 1.00476943e-01,  1.11005591e+03,  2.88831562e-01,\n",
       "          7.48847168e+02,  3.77856255e-01,  2.56556911e+01,\n",
       "          1.88533276e+03,  5.33714867e+00,  7.43940067e+00,\n",
       "          4.80882067e-04,  4.77626705e+00, -1.17032036e-01,\n",
       "          6.87863052e-01],\n",
       "        [ 3.94452721e-01,  1.04785559e+03,  3.72241497e-01,\n",
       "          8.17848511e+02,  6.78906560e-01,  2.02883625e+01,\n",
       "          2.22353833e+03,  3.41546512e+00,  4.46876144e+00,\n",
       "          1.58249994e-03,  3.10951424e+00, -1.57356724e-01,\n",
       "          6.48519218e-01],\n",
       "        [ 5.72002470e-01,  1.56240234e+02,  6.17526174e-01,\n",
       "          1.99688599e+02,  7.95829833e-01,  8.01902199e+00,\n",
       "          6.42514160e+02,  2.28410411e+00,  3.06784320e+00,\n",
       "          3.33046191e-03,  1.96630359e+00, -2.15135306e-01,\n",
       "          7.01621890e-01],\n",
       "        [ 6.17470801e-01,  5.65236694e+02,  4.99952048e-01,\n",
       "          5.45296387e+02,  8.02884042e-01,  1.34302979e+01,\n",
       "          1.61594873e+03,  2.29190278e+00,  2.98895645e+00,\n",
       "          2.41654483e-03,  2.07826924e+00, -2.01190740e-01,\n",
       "          6.57023311e-01],\n",
       "        [ 5.95475286e-02,  2.73832764e+03,  2.49816403e-01,\n",
       "          1.79265271e+03,  2.83715487e-01,  4.40067291e+01,\n",
       "          4.43228271e+03,  6.16110182e+00,  8.38518524e+00,\n",
       "          2.90739787e-04,  5.58953524e+00, -1.36216193e-01,\n",
       "          7.91780889e-01],\n",
       "        [ 1.21245660e-01,  4.34680322e+03,  1.97314635e-01,\n",
       "          2.72808984e+03,  3.64116728e-01,  6.45030594e+01,\n",
       "          6.56555615e+03,  6.15003061e+00,  8.24102974e+00,\n",
       "          4.93683619e-04,  5.56267500e+00, -1.31312057e-01,\n",
       "          7.11884737e-01],\n",
       "        [ 2.83139646e-02,  3.69185278e+03,  1.66726619e-01,\n",
       "          2.13970190e+03,  1.61875278e-01,  7.22402649e+01,\n",
       "          4.86695459e+03,  7.21760941e+00,  1.01896362e+01,\n",
       "          1.39714015e-04,  6.40383577e+00, -1.25659361e-01,\n",
       "          8.15256536e-01],\n",
       "        [ 1.03866935e-01,  9.62047058e+02,  3.10444534e-01,\n",
       "          6.77676147e+02,  3.63918751e-01,  2.41256332e+01,\n",
       "          1.74865759e+03,  5.31223536e+00,  7.15869474e+00,\n",
       "          4.63173579e-04,  4.80791235e+00, -1.12740643e-01,\n",
       "          6.88982129e-01],\n",
       "        [ 1.47100121e-01,  2.09595630e+03,  2.17171744e-01,\n",
       "          1.42879053e+03,  4.52681243e-01,  3.15832253e+01,\n",
       "          3.61920581e+03,  5.05607128e+00,  6.60935163e+00,\n",
       "          6.42108731e-04,  4.59327745e+00, -1.29220486e-01,\n",
       "          7.02665746e-01],\n",
       "        [ 8.16406906e-02,  3.40831616e+03,  2.11127698e-01,\n",
       "          2.07950806e+03,  3.36308271e-01,  5.30075188e+01,\n",
       "          4.90971582e+03,  6.19748259e+00,  8.81536674e+00,\n",
       "          3.66703287e-04,  5.59548855e+00, -1.19143382e-01,\n",
       "          7.48695016e-01],\n",
       "        [ 4.44089502e-01,  1.69798779e+03,  2.87104636e-01,\n",
       "          1.12880920e+03,  6.99608445e-01,  2.34100380e+01,\n",
       "          2.81724902e+03,  3.33684111e+00,  4.62917566e+00,\n",
       "          1.77368929e-03,  3.05169725e+00, -1.95941076e-01,\n",
       "          7.73161173e-01],\n",
       "        [ 5.05587757e-01,  1.06894739e+03,  4.21630919e-01,\n",
       "          8.81585266e+02,  7.37984002e-01,  1.89562664e+01,\n",
       "          2.45739380e+03,  2.95723438e+00,  4.05378723e+00,\n",
       "          1.99697330e-03,  2.66541362e+00, -2.17602253e-01,\n",
       "          7.59026051e-01],\n",
       "        [ 2.55016256e-02,  4.16604053e+03,  2.78876752e-01,\n",
       "          2.80075244e+03,  1.66605517e-01,  8.64153290e+01,\n",
       "          7.03696875e+03,  7.45321035e+00,  1.08710356e+01,\n",
       "          1.28481566e-04,  6.48241615e+00, -1.02879606e-01,\n",
       "          7.53797054e-01],\n",
       "        [ 1.45398960e-01,  2.34123999e+03,  3.76151592e-01,\n",
       "          1.78511157e+03,  4.30472225e-01,  4.39656296e+01,\n",
       "          4.79920605e+03,  5.51296806e+00,  7.94419050e+00,\n",
       "          6.18139165e-04,  4.87566042e+00, -1.34381816e-01,\n",
       "          7.45601118e-01],\n",
       "        [ 3.66013765e-01,  1.07060840e+03,  5.05589128e-01,\n",
       "          1.07017017e+03,  6.36017442e-01,  2.85568428e+01,\n",
       "          3.21007202e+03,  3.89103675e+00,  5.53251410e+00,\n",
       "          1.45398779e-03,  3.42898726e+00, -1.78198174e-01,\n",
       "          7.73243368e-01],\n",
       "        [ 2.38680810e-01,  1.49991211e+03,  3.67649257e-01,\n",
       "          1.17628284e+03,  5.24945498e-01,  2.90536613e+01,\n",
       "          3.20521924e+03,  4.61445236e+00,  6.15409327e+00,\n",
       "          9.58535296e-04,  4.21569109e+00, -1.32097855e-01,\n",
       "          6.92103863e-01],\n",
       "        [ 5.01755893e-01,  1.69544812e+03,  4.76172268e-01,\n",
       "          1.57732642e+03,  7.26988733e-01,  2.97470264e+01,\n",
       "          4.61385791e+03,  3.14007640e+00,  4.40894461e+00,\n",
       "          1.96128967e-03,  2.85530543e+00, -2.33761266e-01,\n",
       "          8.19062650e-01],\n",
       "        [ 3.61860812e-01,  1.38550920e+01,  3.94954294e-01,\n",
       "          1.16090946e+01,  6.87713325e-01,  2.39019060e+00,\n",
       "          3.25812874e+01,  2.49973965e+00,  3.17891073e+00,\n",
       "          5.72664151e-03,  2.23739576e+00, -9.40386653e-02,\n",
       "          4.63619888e-01],\n",
       "        [ 7.64035881e-02,  4.36795996e+03,  3.81764472e-01,\n",
       "          3.32568604e+03,  3.10274065e-01,  6.87055054e+01,\n",
       "          8.93478418e+03,  6.53389168e+00,  9.30695629e+00,\n",
       "          3.37059377e-04,  5.78012800e+00, -1.36857972e-01,\n",
       "          7.98257828e-01],\n",
       "        [ 1.08029312e-02,  6.14186426e+03,  1.89671680e-01,\n",
       "          3.55424194e+03,  1.01816453e-01,  8.26336517e+01,\n",
       "          8.07510400e+03,  7.54990149e+00,  1.14435139e+01,\n",
       "          7.83864816e-05,  6.75548792e+00, -1.16195425e-01,\n",
       "          8.40633273e-01],\n",
       "        [ 4.96356189e-02,  3.52842529e+03,  2.52887398e-01,\n",
       "          2.37958179e+03,  2.71167010e-01,  5.49645157e+01,\n",
       "          5.98990186e+03,  6.53143263e+00,  9.39592457e+00,\n",
       "          2.49624893e-04,  5.83803511e+00, -1.02709077e-01,\n",
       "          6.71258450e-01],\n",
       "        [ 5.01510501e-01,  4.40921211e+01,  3.78924817e-01,\n",
       "          3.62247849e+01,  7.76892364e-01,  2.62504792e+00,\n",
       "          1.00807014e+02,  2.08497548e+00,  2.65028906e+00,\n",
       "          2.57248548e-03,  1.88632834e+00, -9.63330641e-02,\n",
       "          4.25606102e-01],\n",
       "        [ 5.10224521e-01,  1.19195752e+03,  4.82411623e-01,\n",
       "          1.12591272e+03,  7.38140941e-01,  2.14838524e+01,\n",
       "          3.31169336e+03,  2.96166778e+00,  3.93412304e+00,\n",
       "          2.00606231e-03,  2.68710089e+00, -1.91169709e-01,\n",
       "          7.11538851e-01],\n",
       "        [ 2.83469707e-01,  6.89618683e+01,  2.35527188e-01,\n",
       "          4.73300819e+01,  6.04418337e-01,  5.13865328e+00,\n",
       "          1.20358459e+02,  3.22373962e+00,  4.09102297e+00,\n",
       "          2.64741154e-03,  2.94763994e+00, -9.12395716e-02,\n",
       "          4.81121510e-01],\n",
       "        [ 6.14727497e-01,  1.56975378e+03,  3.75205606e-01,\n",
       "          1.20649792e+03,  8.00925434e-01,  1.92636795e+01,\n",
       "          3.25623779e+03,  2.40964508e+00,  3.29016566e+00,\n",
       "          2.41313549e-03,  2.19895697e+00, -2.71963388e-01,\n",
       "          7.87638009e-01],\n",
       "        [ 2.23291650e-01,  9.89398315e+02,  3.04058582e-01,\n",
       "          7.15704224e+02,  5.05356669e-01,  2.60943680e+01,\n",
       "          1.87341870e+03,  4.68890190e+00,  6.27476358e+00,\n",
       "          9.04590474e-04,  4.22916126e+00, -1.13211870e-01,\n",
       "          5.81031919e-01],\n",
       "        [ 6.71480358e-01,  2.67289963e+01,  4.36756819e-01,\n",
       "          2.28544102e+01,  8.76736343e-01,  1.40457225e+00,\n",
       "          6.46886444e+01,  1.37077868e+00,  1.74537694e+00,\n",
       "          3.71088088e-03,  1.15763962e+00, -1.82687476e-01,\n",
       "          5.39128482e-01],\n",
       "        [ 5.28500915e-01,  5.05142456e+02,  4.57495511e-01,\n",
       "          4.52329620e+02,  7.46689081e-01,  1.52168694e+01,\n",
       "          1.30417603e+03,  2.77738905e+00,  3.89613748e+00,\n",
       "          2.07698904e-03,  2.49515915e+00, -1.94041327e-01,\n",
       "          7.30881333e-01],\n",
       "        [ 1.16004474e-01,  1.80628625e+03,  3.04984093e-01,\n",
       "          1.28754297e+03,  3.67919475e-01,  4.16810150e+01,\n",
       "          3.34388574e+03,  5.78702211e+00,  7.94290447e+00,\n",
       "          4.81340656e-04,  5.22191954e+00, -1.03607386e-01,\n",
       "          6.68004870e-01],\n",
       "        [ 1.93612978e-01,  1.05847620e+03,  2.72346944e-01,\n",
       "          6.92028076e+02,  5.08725643e-01,  2.00803928e+01,\n",
       "          1.70963611e+03,  4.46166420e+00,  5.84228230e+00,\n",
       "          9.53645969e-04,  4.04560566e+00, -1.36308163e-01,\n",
       "          6.48139656e-01],\n",
       "        [ 3.30224521e-02,  8.84224792e+02,  3.52038771e-01,\n",
       "          6.69284546e+02,  1.66096285e-01,  3.59486313e+01,\n",
       "          1.79291345e+03,  6.33837986e+00,  8.60761261e+00,\n",
       "          1.80570874e-04,  5.57572174e+00, -1.62493497e-01,\n",
       "          8.77622128e-01],\n",
       "        [ 1.58531949e-01,  1.24219312e+03,  3.67851108e-01,\n",
       "          9.87491272e+02,  4.30231154e-01,  3.04462032e+01,\n",
       "          2.70777197e+03,  5.19459057e+00,  6.91003227e+00,\n",
       "          6.49969326e-04,  4.69685316e+00, -1.14568450e-01,\n",
       "          6.77608311e-01],\n",
       "        [ 1.33758515e-01,  1.83932019e+03,  2.80604929e-01,\n",
       "          1.22270398e+03,  4.00821269e-01,  3.29318123e+01,\n",
       "          3.05149561e+03,  5.40456390e+00,  7.27282906e+00,\n",
       "          5.59311185e-04,  4.93796206e+00, -1.29934788e-01,\n",
       "          7.31302798e-01],\n",
       "        [ 7.12107003e-01,  7.74995117e+01,  3.81424010e-01,\n",
       "          6.59217453e+01,  8.74087334e-01,  2.79158640e+00,\n",
       "          1.86187469e+02,  1.41297305e+00,  1.74922812e+00,\n",
       "          3.99023993e-03,  1.29443824e+00, -1.49586961e-01,\n",
       "          4.44551647e-01],\n",
       "        [ 5.18380225e-01,  2.25969208e+02,  3.82127374e-01,\n",
       "          1.83379074e+02,  7.39013910e-01,  1.03188820e+01,\n",
       "          5.07547089e+02,  2.67951965e+00,  3.80300212e+00,\n",
       "          2.49534822e-03,  2.41411877e+00, -1.98454991e-01,\n",
       "          7.31243491e-01],\n",
       "        [ 1.10806808e-01,  2.26005298e+03,  2.46026143e-01,\n",
       "          1.48410034e+03,  3.73965919e-01,  3.55283318e+01,\n",
       "          3.67634839e+03,  5.56059551e+00,  7.52751255e+00,\n",
       "          4.81943076e-04,  5.08991814e+00, -1.18774265e-01,\n",
       "          6.96240723e-01],\n",
       "        [ 2.28753522e-01,  1.32211105e+02,  3.12768131e-01,\n",
       "          8.97382431e+01,  5.73496699e-01,  6.39098501e+00,\n",
       "          2.26741882e+02,  3.45848441e+00,  4.54730272e+00,\n",
       "          2.24620500e-03,  3.12345433e+00, -9.44589451e-02,\n",
       "          5.31328201e-01],\n",
       "        [ 2.63158232e-01,  1.21152930e+03,  3.22745085e-01,\n",
       "          8.87303772e+02,  5.54563761e-01,  2.48303680e+01,\n",
       "          2.33768579e+03,  4.36072111e+00,  5.63164759e+00,\n",
       "          1.11900212e-03,  3.99222946e+00, -1.29855067e-01,\n",
       "          6.27680361e-01],\n",
       "        [ 3.71654481e-01,  1.41614844e+03,  3.46165955e-01,\n",
       "          1.06864014e+03,  6.37548804e-01,  2.37722206e+01,\n",
       "          2.85841235e+03,  3.75711775e+00,  5.00103235e+00,\n",
       "          1.46992959e-03,  3.44089127e+00, -1.52142987e-01,\n",
       "          6.77794456e-01],\n",
       "        [ 7.08939850e-01,  8.73518616e+02,  3.04517150e-01,\n",
       "          5.93207642e+02,  8.56786251e-01,  1.15301018e+01,\n",
       "          1.49931201e+03,  1.76231647e+00,  2.26256704e+00,\n",
       "          2.77281133e-03,  1.65114284e+00, -2.20138088e-01,\n",
       "          6.25673354e-01],\n",
       "        [ 6.18187547e-01,  2.05477319e+03,  3.54194552e-01,\n",
       "          1.44429395e+03,  8.19536567e-01,  1.77534466e+01,\n",
       "          3.72240259e+03,  2.20751023e+00,  2.90109587e+00,\n",
       "          2.45025079e-03,  1.99835873e+00, -2.85737157e-01,\n",
       "          7.78376758e-01],\n",
       "        [ 9.62186754e-02,  2.70144409e+03,  3.52436006e-01,\n",
       "          2.01449243e+03,  3.44895661e-01,  4.51627579e+01,\n",
       "          5.35652539e+03,  5.96791792e+00,  7.88797188e+00,\n",
       "          4.16428491e-04,  5.39070606e+00, -1.36965245e-01,\n",
       "          7.71527886e-01],\n",
       "        [ 8.96114945e-01,  1.03341698e+02,  3.89961213e-01,\n",
       "          9.04242859e+01,  9.53879237e-01,  2.24027181e+00,\n",
       "          2.58355438e+02,  6.37616873e-01,  7.78190613e-01,\n",
       "          3.70712136e-03,  5.96445858e-01, -2.73837149e-01,\n",
       "          4.58281726e-01],\n",
       "        [ 8.97922218e-02,  2.04621045e+03,  1.61053926e-01,\n",
       "          1.25068921e+03,  3.29319119e-01,  3.52727470e+01,\n",
       "          2.95654663e+03,  5.80532789e+00,  7.65657282e+00,\n",
       "          3.95111245e-04,  5.34884310e+00, -1.21569775e-01,\n",
       "          7.16150463e-01],\n",
       "        [ 2.98890233e-01,  1.59176147e+03,  1.98843285e-01,\n",
       "          1.01745087e+03,  5.70465267e-01,  2.66212254e+01,\n",
       "          2.47804199e+03,  4.27121687e+00,  5.56031322e+00,\n",
       "          1.18308002e-03,  3.97747278e+00, -1.07414268e-01,\n",
       "          5.29745042e-01],\n",
       "        [ 1.94666952e-01,  1.47670728e+03,  3.27573061e-01,\n",
       "          1.04977332e+03,  4.84934717e-01,  2.66542664e+01,\n",
       "          2.72238599e+03,  4.78137827e+00,  6.28450394e+00,\n",
       "          8.23543756e-04,  4.36556244e+00, -1.34308800e-01,\n",
       "          6.89960003e-01],\n",
       "        [ 1.59384221e-01,  2.77762482e+02,  2.67146200e-01,\n",
       "          1.76255310e+02,  4.84136522e-01,  1.02100172e+01,\n",
       "          4.27258759e+02,  4.14444351e+00,  5.54808426e+00,\n",
       "          1.25163188e-03,  3.75218201e+00, -9.15699452e-02,\n",
       "          5.76081574e-01],\n",
       "        [ 2.26969033e-01,  9.96843445e+02,  2.73896307e-01,\n",
       "          6.66879395e+02,  5.21194160e-01,  2.36484699e+01,\n",
       "          1.67067419e+03,  4.55159521e+00,  6.20916748e+00,\n",
       "          9.54547373e-04,  4.13851881e+00, -1.16420157e-01,\n",
       "          6.19105995e-01],\n",
       "        [ 3.88815589e-02,  3.55715771e+03,  1.87160268e-01,\n",
       "          2.11217163e+03,  2.33330920e-01,  5.17535172e+01,\n",
       "          4.89152881e+03,  6.52440405e+00,  9.05968857e+00,\n",
       "          2.10322120e-04,  5.90625954e+00, -1.31480351e-01,\n",
       "          8.05480063e-01],\n",
       "        [ 8.82053375e-02,  3.48369141e+03,  2.57277489e-01,\n",
       "          2.28250024e+03,  3.44038367e-01,  4.61548347e+01,\n",
       "          5.64631006e+03,  5.85647631e+00,  7.94346571e+00,\n",
       "          3.96196148e-04,  5.36004305e+00, -1.28669828e-01,\n",
       "          7.63981402e-01],\n",
       "        [ 5.67453429e-02,  1.66769641e+03,  1.50834426e-01,\n",
       "          9.50951843e+02,  2.96611100e-01,  3.79381599e+01,\n",
       "          2.13611108e+03,  6.06421947e+00,  8.83932209e+00,\n",
       "          2.95300124e-04,  5.45767164e+00, -9.41077247e-02,\n",
       "          6.65952742e-01],\n",
       "        [ 5.13190985e-01,  8.06057922e+02,  4.28466171e-01,\n",
       "          6.83024414e+02,  7.33807087e-01,  1.84735641e+01,\n",
       "          1.92603979e+03,  2.94125867e+00,  3.93103766e+00,\n",
       "          2.05131760e-03,  2.67153811e+00, -1.77843660e-01,\n",
       "          6.80538833e-01],\n",
       "        [ 1.33841947e-01,  1.54315137e+03,  2.84814239e-01,\n",
       "          1.07486816e+03,  3.86978298e-01,  4.28168106e+01,\n",
       "          2.75632153e+03,  5.64666986e+00,  7.99117279e+00,\n",
       "          5.52660786e-04,  5.02828932e+00, -9.85374525e-02,\n",
       "          5.93267381e-01],\n",
       "        [ 3.26307677e-02,  4.26824854e+03,  2.07673386e-01,\n",
       "          2.65239502e+03,  1.93974212e-01,  7.94223633e+01,\n",
       "          6.34133154e+03,  7.19963932e+00,  1.01160688e+01,\n",
       "          1.58059076e-04,  6.38370705e+00, -1.28614590e-01,\n",
       "          8.14648271e-01],\n",
       "        [ 1.70112655e-01,  4.24293274e+02,  2.69048750e-01,\n",
       "          2.77792084e+02,  4.62165028e-01,  1.79748611e+01,\n",
       "          6.86875061e+02,  4.68237305e+00,  6.63130093e+00,\n",
       "          8.30340781e-04,  4.15266085e+00, -1.08857520e-01,\n",
       "          6.12942159e-01],\n",
       "        [ 2.55349278e-01,  7.36697449e+02,  5.42618096e-01,\n",
       "          7.97825806e+02,  5.62125921e-01,  2.31525517e+01,\n",
       "          2.45460571e+03,  4.25597525e+00,  5.74952221e+00,\n",
       "          1.04925234e-03,  3.75928020e+00, -1.39471442e-01,\n",
       "          6.99460149e-01],\n",
       "        [ 8.54654387e-02,  1.23771191e+03,  1.66017249e-01,\n",
       "          7.34299072e+02,  3.18730950e-01,  3.18599262e+01,\n",
       "          1.69948438e+03,  5.76546574e+00,  7.87765837e+00,\n",
       "          3.84353509e-04,  5.27043629e+00, -1.02696829e-01,\n",
       "          7.02816904e-01],\n",
       "        [ 4.68141496e-01,  1.66397473e+03,  3.13879669e-01,\n",
       "          1.11809167e+03,  7.07368314e-01,  2.27935429e+01,\n",
       "          2.80839209e+03,  3.22340560e+00,  4.44488859e+00,\n",
       "          1.85065693e-03,  2.93739009e+00, -2.10112482e-01,\n",
       "          7.65876114e-01],\n",
       "        [ 3.79439518e-02,  3.13753809e+03,  1.58814386e-01,\n",
       "          1.84103918e+03,  2.56079048e-01,  5.35919304e+01,\n",
       "          4.22661816e+03,  6.53024626e+00,  9.60242081e+00,\n",
       "          2.08516576e-04,  5.90793371e+00, -8.09745044e-02,\n",
       "          6.60682082e-01],\n",
       "        [ 8.66650790e-02,  2.33293115e+03,  4.73008603e-01,\n",
       "          2.16423291e+03,  3.37159812e-01,  5.17795677e+01,\n",
       "          6.32400000e+03,  6.11343336e+00,  8.32371807e+00,\n",
       "          3.94197152e-04,  5.34498596e+00, -1.59461573e-01,\n",
       "          8.14384997e-01],\n",
       "        [ 1.02566518e-01,  1.88079858e+03,  2.93947756e-01,\n",
       "          1.31466736e+03,  3.72611612e-01,  3.41686745e+01,\n",
       "          3.37787085e+03,  5.53591824e+00,  7.44045830e+00,\n",
       "          4.58850525e-04,  5.01663351e+00, -1.16188295e-01,\n",
       "          6.95528030e-01],\n",
       "        [ 1.45308360e-01,  1.66509680e+03,  2.99506009e-01,\n",
       "          1.21563989e+03,  4.09290463e-01,  4.23609962e+01,\n",
       "          3.19746289e+03,  5.55325937e+00,  7.54458761e+00,\n",
       "          6.30915747e-04,  4.95070219e+00, -1.09895155e-01,\n",
       "          6.15452707e-01]], dtype=float32), array([[1, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 0, 0, 0],\n",
       "        ...,\n",
       "        [1, 0, 0, ..., 1, 0, 0],\n",
       "        [1, 0, 0, ..., 0, 0, 0],\n",
       "        [0, 0, 0, ..., 1, 0, 0]]))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "next(genrtor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.losses import binary_crossentropy\n",
    "from keras.optimizers import Adadelta\n",
    "from keras.initializers import VarianceScaling\n",
    "\n",
    "class FeatureBaseLineModel:\n",
    "    \n",
    "    def __init__(self, modelparameter, metrics):\n",
    "        self.params = modelparameter\n",
    "        self.use_multi_gpu = self.params.use_multi_gpu\n",
    "        self.num_classes = self.params.num_classes\n",
    "        self.img_rows = self.params.scaled_row_dim\n",
    "        self.img_cols = self.params.scaled_col_dim\n",
    "        self.n_channels = self.params.n_channels\n",
    "        self.input_shape = (13,)\n",
    "        self.my_metrics = metrics\n",
    "    \n",
    "    def build_model(self):\n",
    "        self.model = Sequential()\n",
    "        self.model.add(Dense(64, kernel_initializer=VarianceScaling(seed=0), activation='relu', input_shape=self.input_shape))\n",
    "        self.model.add(Dense(13, kernel_initializer=VarianceScaling(seed=0), activation='relu'))\n",
    "        self.model.add(Dense(5, kernel_initializer=VarianceScaling(seed=0), activation='relu'))\n",
    "        self.model.add(Dense(self.num_classes, activation='sigmoid'))\n",
    "    \n",
    "    def compile_model(self, loss=keras.losses.categorical_crossentropy, optimizer=keras.optimizers.Adadelta()):\n",
    "        self.model.compile(loss=loss,\n",
    "              optimizer=optimizer,\n",
    "              metrics=self.my_metrics)\n",
    "    \n",
    "    def set_generators(self, train_generator, validation_generator, tr_sz, vl_sz):\n",
    "        self.training_generator = train_generator\n",
    "        self.validation_generator = validation_generator\n",
    "        self.tr_sz = tr_sz\n",
    "        self.vl_sz = vl_sz\n",
    "    \n",
    "    def learn(self, use_multiprocessing=False, workers= 24):\n",
    "        return self.model.fit_generator(generator=self.training_generator,\n",
    "                    validation_data=self.validation_generator,\n",
    "                    epochs=self.params.n_epochs, \n",
    "                    use_multiprocessing=use_multiprocessing,\n",
    "                    workers=workers, verbose=1, steps_per_epoch=self.tr_sz//self.params.batch_size, validation_steps=self.vl_sz//self.params.batch_size)\n",
    "    \n",
    "    def score(self, use_multiprocessing=False, workers=24):\n",
    "        return self.model.evaluate_generator(generator=self.validation_generator,\n",
    "                                      use_multiprocessing=use_multiprocessing, \n",
    "                                      workers=workers)\n",
    "    \n",
    "    def predict(self, predict_generator):\n",
    "        y = predict_generator.predict(self.model)\n",
    "        return y\n",
    "    \n",
    "    def save(self, modeloutputpath):\n",
    "        self.model.save(modeloutputpath)\n",
    "    \n",
    "    def load(self, modelinputpath):\n",
    "        self.model = load_model(modelinputpath)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_5 (Dense)              (None, 64)                896       \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 13)                845       \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 5)                 70        \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 28)                168       \n",
      "=================================================================\n",
      "Total params: 1,979\n",
      "Trainable params: 1,979\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model_feature = FeatureBaseLineModel(parameter, ['accuracy'])\n",
    "model_feature.build_model()\n",
    "model_feature.compile_model()\n",
    "model_feature.model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from ipdb import set_trace\n",
    "from generators import PredictGenerator\n",
    "def train_feature_model(model, partitions, train_labels, train_path, parameter, preprocessor):\n",
    "    target_names = train_labels.drop([\"Target\", \"Id\"], axis=1).columns\n",
    "    \n",
    "    predictions = []\n",
    "    histories = []\n",
    "    \n",
    "    for i, partition in enumerate(partitions):\n",
    "        print(\"training in partition \",i+1)\n",
    "        \n",
    "        training_generator = generator(partition['train'], train_labels, parameter, preprocessor)\n",
    "        tr_sz = partition['train'].shape[0]\n",
    "        vl_sz = partition['validation'].shape[0]\n",
    "        validation_generator = generator(partition['validation'], train_labels, parameter, preprocessor)\n",
    "        predict_generator = PredictGenerator(partition['validation'], preprocessor, train_path)\n",
    "        model.set_generators(training_generator, validation_generator, tr_sz, vl_sz)\n",
    "        histories.append(model.learn())\n",
    "        \n",
    "        proba_predictions = model.predict(predict_generator)\n",
    "        proba_predictions = pd.DataFrame(index = partition['validation'], data=proba_predictions, columns=target_names)\n",
    "        \n",
    "        predictions.append(proba_predictions)\n",
    "        \n",
    "    return predictions, histories"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training in partition  1\n",
      "Epoch 1/5\n"
     ]
    }
   ],
   "source": [
    "pred_feature_model, hist_feature_model = train_feature_model(model_feature, partitions, train_labels, train_path, parameter, preprocessor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.2.4'"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras.__version__"
   ]
  },
  {
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}