Model_00_DAug.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from os.path import join as opj\n",
    "from matplotlib import pyplot as plt\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "import pylab\n",
    "plt.rcParams['figure.figsize'] = 10, 10\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Load the data.\n",
    "train = pd.read_json(\"input/train.json\")\n",
    "target_train = np.array(train['is_iceberg'])\n",
    "test = pd.read_json(\"input/test.json\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Generate the training data\n",
    "#Create 3 bands having HH, HV and avg of both\n",
    "X_band_1=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train[\"band_1\"]])\n",
    "X_band_2=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train[\"band_2\"]])\n",
    "X_train = np.concatenate([X_band_1[:, :, :, np.newaxis], X_band_2[:, :, :, np.newaxis],((X_band_1+X_band_2)/2)[:, :, :, np.newaxis]], axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.inc_angle = train.inc_angle.replace('na', 0)\n",
    "idx_tr = np.where(train.inc_angle > 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_train = target_train[idx_tr[0]]\n",
    "X_train = X_train[idx_tr[0],...]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import cv2\n",
    "# def get_more_images(imgs):\n",
    "    \n",
    "#     more_images = []\n",
    "#     vert_flip_imgs = []\n",
    "#     hori_flip_imgs = []\n",
    "      \n",
    "#     for i in range(0,imgs.shape[0]):\n",
    "#         a=imgs[i,:,:,0]\n",
    "#         b=imgs[i,:,:,1]\n",
    "#         c=imgs[i,:,:,2]\n",
    "        \n",
    "#         av=cv2.flip(a,1)\n",
    "#         ah=cv2.flip(a,0)\n",
    "#         bv=cv2.flip(b,1)\n",
    "#         bh=cv2.flip(b,0)\n",
    "#         cv=cv2.flip(c,1)\n",
    "#         ch=cv2.flip(c,0)\n",
    "        \n",
    "#         vert_flip_imgs.append(np.dstack((av, bv, cv)))\n",
    "#         hori_flip_imgs.append(np.dstack((ah, bh, ch)))\n",
    "      \n",
    "#     v = np.array(vert_flip_imgs)\n",
    "#     h = np.array(hori_flip_imgs)\n",
    "       \n",
    "#     more_images = np.concatenate((imgs,v,h))\n",
    "    \n",
    "#     return more_images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# X_train_more = get_more_images(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Y_train_more = np.concatenate((Y_train, Y_train, Y_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "#Import Keras.\n",
    "from matplotlib import pyplot\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D, Dense, Dropout, Input, Flatten, Activation\n",
    "from keras.layers import GlobalMaxPooling2D\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.merge import Concatenate\n",
    "from keras.models import Model\n",
    "from keras import initializers\n",
    "from keras.optimizers import Adam\n",
    "from keras.callbacks import ModelCheckpoint, Callback, EarlyStopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Data Aug for multi-input\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "batch_size=64\n",
    "# Define the image transformations here\n",
    "gen = ImageDataGenerator(horizontal_flip = True,\n",
    "                         vertical_flip = True,\n",
    "                         width_shift_range = 0.,\n",
    "                         height_shift_range = 0.,\n",
    "                         channel_shift_range=0,\n",
    "                         zoom_range = 0.2,\n",
    "                         rotation_range = 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# X_batch, y_batch = gen.flow(X_train, X_train, batch_size=batch_size)\n",
    "# def gen_flow_for_two_inputs(X, y):\n",
    "#     genX1 = gen.flow(X,y,  batch_size=batch_size,seed=55)\n",
    "#     genX2 = gen.flow(X1,X2, batch_size=batch_size,seed=55)\n",
    "#     while True:\n",
    "#             X1i = genX1.next()\n",
    "#             X2i = genX2.next()\n",
    "#             #Assert arrays are equal - this was for peace of mind, but slows down training\n",
    "#             #np.testing.assert_array_equal(X1i[0],X2i[0])\n",
    "#             yield [X1i[0], X2i[1]], X1i[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "#define our model\n",
    "def getModel():\n",
    "    #Building the model\n",
    "    model=Sequential()\n",
    "    model.add(Conv2D(32, 3, activation=\"relu\", input_shape=(75, 75, 3)))\n",
    "    model.add(Dropout(0.2))\n",
    "    model.add(MaxPooling2D())\n",
    "\n",
    "    model.add(Conv2D(64, 3, activation=\"relu\", input_shape=(75, 75, 3)))\n",
    "    model.add(Dropout(0.2))\n",
    "    model.add(MaxPooling2D())\n",
    "\n",
    "    model.add(GlobalMaxPooling2D())\n",
    "    model.add(Dropout(0.3))\n",
    "    model.add(Dense(1, activation=\"sigmoid\"))\n",
    "    \n",
    "    adam=Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)\n",
    "    model.compile(adam, \"binary_crossentropy\", metrics=[\"accuracy\"])\n",
    "    \n",
    "    model.summary()\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_callbacks(filepath, patience=2):\n",
    "    es = EarlyStopping('val_loss', patience=patience, mode=\"min\")\n",
    "    msave = ModelCheckpoint(filepath, save_best_only=True)\n",
    "    return [es, msave]\n",
    "file_path = \".model_weights.hdf5\"\n",
    "callbacks = get_callbacks(filepath=file_path, patience=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_cv, X_valid, y_train_cv, y_valid = train_test_split(X_train, Y_train, random_state=1, train_size=0.75)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_9 (Conv2D)            (None, 73, 73, 32)        896       \n",
      "_________________________________________________________________\n",
      "dropout_13 (Dropout)         (None, 73, 73, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_9 (MaxPooling2 (None, 36, 36, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_10 (Conv2D)           (None, 34, 34, 64)        18496     \n",
      "_________________________________________________________________\n",
      "dropout_14 (Dropout)         (None, 34, 34, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_10 (MaxPooling (None, 17, 17, 64)        0         \n",
      "_________________________________________________________________\n",
      "global_max_pooling2d_5 (Glob (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_15 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 1)                 65        \n",
      "=================================================================\n",
      "Total params: 19,457\n",
      "Trainable params: 19,457\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/100\n",
      "24/24 [==============================] - 1s 45ms/step - loss: 2.0187 - acc: 0.6004 - val_loss: 0.5239 - val_acc: 0.7391\n",
      "Epoch 2/100\n",
      "24/24 [==============================] - 1s 42ms/step - loss: 0.7721 - acc: 0.6628 - val_loss: 0.4859 - val_acc: 0.7473\n",
      "Epoch 3/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.5102 - acc: 0.7337 - val_loss: 0.4709 - val_acc: 0.8016\n",
      "Epoch 4/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4636 - acc: 0.7768 - val_loss: 0.4468 - val_acc: 0.8179\n",
      "Epoch 5/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.4581 - acc: 0.7708 - val_loss: 0.4544 - val_acc: 0.8370\n",
      "Epoch 6/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.4600 - acc: 0.7848 - val_loss: 0.4606 - val_acc: 0.8261\n",
      "Epoch 7/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.4729 - acc: 0.7708 - val_loss: 0.4500 - val_acc: 0.8451\n",
      "Epoch 8/100\n",
      "24/24 [==============================] - 1s 36ms/step - loss: 0.4470 - acc: 0.7982 - val_loss: 0.4517 - val_acc: 0.8424\n",
      "Epoch 9/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.4456 - acc: 0.7930 - val_loss: 0.4152 - val_acc: 0.7935\n",
      "Epoch 10/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4608 - acc: 0.7781 - val_loss: 0.4181 - val_acc: 0.8261\n",
      "Epoch 11/100\n",
      "24/24 [==============================] - 1s 36ms/step - loss: 0.4013 - acc: 0.8280 - val_loss: 0.4271 - val_acc: 0.8179\n",
      "Epoch 12/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.4666 - acc: 0.7721 - val_loss: 0.4068 - val_acc: 0.8261\n",
      "Epoch 13/100\n",
      "24/24 [==============================] - 1s 36ms/step - loss: 0.4141 - acc: 0.8030 - val_loss: 0.4231 - val_acc: 0.8261\n",
      "Epoch 14/100\n",
      "24/24 [==============================] - 1s 41ms/step - loss: 0.4118 - acc: 0.8125 - val_loss: 0.3912 - val_acc: 0.8288\n",
      "Epoch 15/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.4495 - acc: 0.7769 - val_loss: 0.4188 - val_acc: 0.8424\n",
      "Epoch 16/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4040 - acc: 0.8146 - val_loss: 0.3725 - val_acc: 0.8397\n",
      "Epoch 17/100\n",
      "24/24 [==============================] - 1s 41ms/step - loss: 0.4320 - acc: 0.7917 - val_loss: 0.4050 - val_acc: 0.8125\n",
      "Epoch 18/100\n",
      "24/24 [==============================] - 1s 41ms/step - loss: 0.4131 - acc: 0.7899 - val_loss: 0.4144 - val_acc: 0.8342\n",
      "Epoch 19/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.4084 - acc: 0.8185 - val_loss: 0.3677 - val_acc: 0.8451\n",
      "Epoch 20/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.4064 - acc: 0.8146 - val_loss: 0.3719 - val_acc: 0.8560\n",
      "Epoch 21/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.4184 - acc: 0.7995 - val_loss: 0.4325 - val_acc: 0.8043\n",
      "Epoch 22/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4223 - acc: 0.8057 - val_loss: 0.3948 - val_acc: 0.8125\n",
      "Epoch 23/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4435 - acc: 0.8029 - val_loss: 0.3728 - val_acc: 0.8451\n",
      "Epoch 24/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3912 - acc: 0.8198 - val_loss: 0.3579 - val_acc: 0.8533\n",
      "Epoch 25/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.4110 - acc: 0.8216 - val_loss: 0.3665 - val_acc: 0.8560\n",
      "Epoch 26/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4242 - acc: 0.8122 - val_loss: 0.3781 - val_acc: 0.8478\n",
      "Epoch 27/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3826 - acc: 0.8241 - val_loss: 0.3586 - val_acc: 0.8533\n",
      "Epoch 28/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3942 - acc: 0.8307 - val_loss: 0.3675 - val_acc: 0.8587\n",
      "Epoch 29/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3765 - acc: 0.8146 - val_loss: 0.3612 - val_acc: 0.8641\n",
      "Epoch 30/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.4145 - acc: 0.8086 - val_loss: 0.3529 - val_acc: 0.8560\n",
      "Epoch 31/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3772 - acc: 0.8238 - val_loss: 0.3742 - val_acc: 0.8668\n",
      "Epoch 32/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3564 - acc: 0.8473 - val_loss: 0.3453 - val_acc: 0.8587\n",
      "Epoch 33/100\n",
      "24/24 [==============================] - 1s 41ms/step - loss: 0.3788 - acc: 0.8289 - val_loss: 0.3624 - val_acc: 0.8587\n",
      "Epoch 34/100\n",
      "24/24 [==============================] - 1s 40ms/step - loss: 0.3665 - acc: 0.8359 - val_loss: 0.3541 - val_acc: 0.8641\n",
      "Epoch 35/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3884 - acc: 0.8159 - val_loss: 0.3561 - val_acc: 0.8397\n",
      "Epoch 36/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.3897 - acc: 0.8237 - val_loss: 0.4105 - val_acc: 0.8505\n",
      "Epoch 37/100\n",
      "24/24 [==============================] - 1s 40ms/step - loss: 0.3917 - acc: 0.8302 - val_loss: 0.3755 - val_acc: 0.8505\n",
      "Epoch 38/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3779 - acc: 0.8303 - val_loss: 0.3376 - val_acc: 0.8587\n",
      "Epoch 39/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3656 - acc: 0.8346 - val_loss: 0.3336 - val_acc: 0.8505\n",
      "Epoch 40/100\n",
      "24/24 [==============================] - 1s 36ms/step - loss: 0.3901 - acc: 0.8196 - val_loss: 0.3560 - val_acc: 0.8533\n",
      "Epoch 41/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.3799 - acc: 0.8177 - val_loss: 0.3368 - val_acc: 0.8478\n",
      "Epoch 42/100\n",
      "24/24 [==============================] - 1s 36ms/step - loss: 0.3645 - acc: 0.8303 - val_loss: 0.3595 - val_acc: 0.8505\n",
      "Epoch 43/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3730 - acc: 0.8255 - val_loss: 0.3288 - val_acc: 0.8614\n",
      "Epoch 44/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3889 - acc: 0.8277 - val_loss: 0.3874 - val_acc: 0.8288\n",
      "Epoch 45/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3762 - acc: 0.8314 - val_loss: 0.3579 - val_acc: 0.8696\n",
      "Epoch 46/100\n",
      "24/24 [==============================] - 1s 41ms/step - loss: 0.3630 - acc: 0.8289 - val_loss: 0.3396 - val_acc: 0.8641\n",
      "Epoch 47/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.3664 - acc: 0.8411 - val_loss: 0.3301 - val_acc: 0.8723\n",
      "Epoch 48/100\n",
      "24/24 [==============================] - 1s 36ms/step - loss: 0.3779 - acc: 0.8184 - val_loss: 0.3734 - val_acc: 0.8723\n",
      "Epoch 49/100\n",
      "24/24 [==============================] - 1s 48ms/step - loss: 0.3663 - acc: 0.8151 - val_loss: 0.3202 - val_acc: 0.8641\n",
      "Epoch 50/100\n",
      "24/24 [==============================] - 1s 34ms/step - loss: 0.3461 - acc: 0.8402 - val_loss: 0.3322 - val_acc: 0.8424\n",
      "Epoch 51/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3723 - acc: 0.8359 - val_loss: 0.3475 - val_acc: 0.8424\n",
      "Epoch 52/100\n",
      "24/24 [==============================] - 1s 42ms/step - loss: 0.3476 - acc: 0.8486 - val_loss: 0.3213 - val_acc: 0.8587\n",
      "Epoch 53/100\n",
      "24/24 [==============================] - 1s 41ms/step - loss: 0.3419 - acc: 0.8366 - val_loss: 0.3125 - val_acc: 0.8560\n",
      "Epoch 54/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3992 - acc: 0.8060 - val_loss: 0.3539 - val_acc: 0.8560\n",
      "Epoch 55/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3691 - acc: 0.8289 - val_loss: 0.3243 - val_acc: 0.8750\n",
      "Epoch 56/100\n",
      "24/24 [==============================] - 1s 42ms/step - loss: 0.3722 - acc: 0.8315 - val_loss: 0.3169 - val_acc: 0.8723\n",
      "Epoch 57/100\n",
      "24/24 [==============================] - 1s 39ms/step - loss: 0.3635 - acc: 0.8255 - val_loss: 0.3500 - val_acc: 0.8451\n",
      "Epoch 58/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.4012 - acc: 0.8112 - val_loss: 0.3384 - val_acc: 0.8641\n",
      "Epoch 59/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3917 - acc: 0.8115 - val_loss: 0.4431 - val_acc: 0.8478\n",
      "Epoch 60/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3839 - acc: 0.8411 - val_loss: 0.3430 - val_acc: 0.8668\n",
      "Epoch 61/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3708 - acc: 0.8289 - val_loss: 0.3347 - val_acc: 0.8723\n",
      "Epoch 62/100\n",
      "24/24 [==============================] - 1s 38ms/step - loss: 0.3570 - acc: 0.8320 - val_loss: 0.3306 - val_acc: 0.8505\n",
      "Epoch 63/100\n",
      "24/24 [==============================] - 1s 37ms/step - loss: 0.3964 - acc: 0.8257 - val_loss: 0.3250 - val_acc: 0.8397\n"
     ]
    }
   ],
   "source": [
    "model=getModel()\n",
    "\n",
    "# histr = model.fit(X_train_cv, y_train_cv,\n",
    "#           batch_size=24,\n",
    "#           epochs=50,\n",
    "#           verbose=1,\n",
    "#           validation_data=(X_valid, y_valid),\n",
    "#           callbacks=callbacks)\n",
    "\n",
    "gen_flow = gen.flow(X_train_cv, y_train_cv)\n",
    "histr = model.fit_generator(\n",
    "            gen_flow,\n",
    "            steps_per_epoch=24,\n",
    "            epochs=100,\n",
    "            shuffle=True,\n",
    "            verbose=1,\n",
    "            validation_data=(X_valid, y_valid),\n",
    "            callbacks=callbacks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1103/1103 [==============================] - 0s 254us/step\n",
      "Train score: 0.317835671785\n",
      "Train accuracy: 0.855847689042\n"
     ]
    }
   ],
   "source": [
    "model.load_weights(filepath=file_path)\n",
    "score = model.evaluate(X_train_cv, y_train_cv, verbose=1)\n",
    "print('Train score:', score[0])\n",
    "print('Train accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "368/368 [==============================] - 0s 244us/step\n",
      "Test score: 0.3125035698\n",
      "Test accuracy: 0.85597826087\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(X_valid, y_valid, verbose=1)\n",
    "print('Test score:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_band_test_1=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in test[\"band_1\"]])\n",
    "X_band_test_2=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in test[\"band_2\"]])\n",
    "X_test = np.concatenate([X_band_test_1[:, :, :, np.newaxis]\n",
    "                          , X_band_test_2[:, :, :, np.newaxis]\n",
    "                         , ((X_band_test_1+X_band_test_2)/2)[:, :, :, np.newaxis]], axis=-1)\n",
    "predicted_test=model.predict_proba(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "submission = pd.DataFrame()\n",
    "submission['id']=test['id']\n",
    "submission['is_iceberg']=predicted_test.reshape((predicted_test.shape[0]))\n",
    "submission.to_csv('Model_00-DAug.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['acc', 'loss', 'val_loss', 'val_acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fee7f0bab38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEWCAYAAAB1xKBvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fee7ef4c278>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# list all data in history\n",
    "print(histr.history.keys())\n",
    "# summarize history for accuracy\n",
    "plt.plot(histr.history['acc'])\n",
    "plt.plot(histr.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "# summarize history for loss\n",
    "plt.plot(histr.history['loss'])\n",
    "plt.plot(histr.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}