happy-house.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/paperspace/anaconda3/envs/fastai/lib/python3.6/site-packages/sklearn/ensemble/weight_boosting.py:29: DeprecationWarning: numpy.core.umath_tests is an internal NumPy module and should not be imported. It will be removed in a future NumPy release.\n",
      "  from numpy.core.umath_tests import inner1d\n"
     ]
    }
   ],
   "source": [
    "from fastai.conv_learner import *\n",
    "import h5py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True True\n"
     ]
    }
   ],
   "source": [
    "PATH = 'data/happy-house/'\n",
    "os.makedirs('data/happy-house/models', exist_ok=True)\n",
    "os.makedirs('cache/happy-house/tmp', exist_ok=True)\n",
    "print(torch.cuda.is_available(), torch.backends.cudnn.enabled)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "sz = 48\n",
    "arch=resnext50\n",
    "# data set downloaded from kaggle competition - https://www.kaggle.com/iarunava/happy-house-dataset\n",
    "def load_dataset():\n",
    "    train_data = h5py.File('data/happy-house/train_happy.h5', \"r\")\n",
    "    x_train = np.array(train_data[\"train_set_x\"][:]) \n",
    "    y_train = np.array(train_data[\"train_set_y\"][:]) \n",
    "\n",
    "    test_data = h5py.File('data/happy-house/test_happy.h5', \"r\")\n",
    "    x_test = np.array(test_data[\"test_set_x\"][:])\n",
    "    y_test = np.array(test_data[\"test_set_y\"][:]) \n",
    "    \n",
    "    y_train = y_train.reshape((1, y_train.shape[0]))\n",
    "    y_test = y_test.reshape((1, y_test.shape[0]))\n",
    "    \n",
    "    return x_train, y_train, x_test, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, Y_train, X_test, Y_test = load_dataset()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_train = Y_train.T.squeeze()\n",
    "Y_test = Y_test.T.squeeze()\n",
    "def get_data(sz):\n",
    "    tfms = tfms_from_model(arch, sz, aug_tfms=transforms_side_on, max_zoom=1.1)\n",
    "    return ImageClassifierData.from_arrays(\n",
    "        path=PATH,\n",
    "        trn=(X_train, Y_train),\n",
    "        val=(X_test, Y_test),\n",
    "        bs=30,\n",
    "        classes=Y_train,\n",
    "        tfms=tfms)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = get_data(sz)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = ConvLearner.pretrained(arch, data, precompute=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f1d7bf02919e41a0b3f46bceb694724b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=1, style=ProgressStyle(description_width='initial…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 1/20 [00:00<00:14,  1.31it/s, loss=1.07]\n",
      "                                                         \r"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# finding optimal learning rate\n",
    "lrf = learn.lr_find()\n",
    "learn.sched.plot_lr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f12a14efa6b0489ca4b25b89cb5a1117",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=121, style=ProgressStyle(description_width='initi…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                 \n",
      "    0      0.600742   0.51059    0.8       \n",
      "    1      0.628339   0.378149   0.893333                 \n",
      "    2      0.50211    0.261331   0.94                     \n",
      "    3      0.404538   0.193078   0.933333                 \n",
      "    4      0.41436    0.259025   0.92                     \n",
      "    5      0.389685   0.201744   0.9                        \n",
      "    6      0.353408   0.231291   0.933333                 \n",
      "    7      0.31348    0.134138   0.953333                 \n",
      "    8      0.279238   0.177227   0.92                     \n",
      "    9      0.248327   0.11978    0.966667                 \n",
      "    10     0.215498   0.120035   0.96                     \n",
      "    11     0.194384   0.123963   0.966667                 \n",
      "    12     0.175264   0.119691   0.966667                 \n",
      "    13     0.207676   0.13535    0.96                     \n",
      "    14     0.256194   0.349808   0.846667                 \n",
      "    15     0.24524    0.279233   0.906667                 \n",
      "    16     0.258022   0.339209   0.906667                 \n",
      "    17     0.285594   0.132465   0.966667                 \n",
      "    18     0.299753   0.607397   0.873333                \n",
      "    19     0.324062   0.157595   0.933333                 \n",
      "    20     0.367942   0.300846   0.94                     \n",
      "    21     0.344573   0.192059   0.953333                \n",
      "    22     0.296089   0.177619   0.946667                 \n",
      "    23     0.287686   0.144688   0.953333                   \n",
      "    24     0.265064   0.13446    0.966667                 \n",
      "    25     0.224064   0.143392   0.933333                 \n",
      "    26     0.2252     0.142043   0.96                     \n",
      "    27     0.206838   0.124049   0.96                     \n",
      "    28     0.178662   0.108039   0.966667                 \n",
      "    29     0.153946   0.135858   0.973333                 \n",
      "    30     0.146772   0.142713   0.94                     \n",
      "    31     0.132277   0.113583   0.966667                 \n",
      "    32     0.142147   0.112539   0.953333                 \n",
      "    33     0.128267   0.111747   0.953333                 \n",
      "    34     0.137444   0.115881   0.966667                 \n",
      "    35     0.129734   0.10629    0.966667                 \n",
      "    36     0.109309   0.102058   0.966667                 \n",
      "    37     0.097025   0.100459   0.966667                 \n",
      "    38     0.09362    0.104617   0.966667                  \n",
      "    39     0.081504   0.103173   0.966667                  \n",
      "    40     0.118654   0.139336   0.946667                 \n",
      "    41     0.116286   0.146259   0.94                     \n",
      "    42     0.119949   0.144782   0.96                     \n",
      "    43     0.127859   0.150488   0.96                     \n",
      "    44     0.150053   0.113951   0.973333                 \n",
      "    45     0.162807   0.119021   0.966667                 \n",
      "    46     0.151581   0.106105   0.966667                 \n",
      "    47     0.152702   0.130884   0.966667                 \n",
      "    48     0.1526     0.102166   0.973333                 \n",
      "    49     0.147438   0.126056   0.96                    \n",
      "    50     0.138683   0.112741   0.973333                 \n",
      "    51     0.116679   0.132242   0.946667                 \n",
      "    52     0.117821   0.158031   0.953333                \n",
      "    53     0.11899    0.149092   0.946667                 \n",
      "    54     0.129924   0.105254   0.96                     \n",
      "    55     0.14038    0.183997   0.926667                 \n",
      "    56     0.137921   0.092017   0.96                      \n",
      "    57     0.124432   0.101005   0.966667                 \n",
      "    58     0.115304   0.112513   0.966667                 \n",
      "    59     0.109473   0.113973   0.966667                 \n",
      "    60     0.09438    0.135529   0.96                     \n",
      "    61     0.095374   0.130016   0.94                      \n",
      "    62     0.093619   0.099366   0.966667                 \n",
      "    63     0.089786   0.098724   0.973333                  \n",
      "    64     0.087885   0.140388   0.96                      \n",
      "    65     0.091693   0.095396   0.973333                  \n",
      "    66     0.095467   0.128215   0.953333                  \n",
      "    67     0.08382    0.101094   0.973333                    \n",
      "    68     0.079691   0.096137   0.973333                  \n",
      "    69     0.066075   0.099228   0.973333                  \n",
      "    70     0.0756     0.090894   0.973333                  \n",
      "    71     0.08434    0.088699   0.973333                  \n",
      "    72     0.08785    0.111667   0.966667                  \n",
      "    73     0.096944   0.108658   0.966667                  \n",
      "    74     0.088003   0.088863   0.98                      \n",
      "    75     0.070761   0.084842   0.973333                  \n",
      "    76     0.088113   0.107419   0.973333                  \n",
      "    77     0.081242   0.083852   0.98                      \n",
      "    78     0.074769   0.088494   0.98                      \n",
      "    79     0.080027   0.06911    0.973333                  \n",
      "    80     0.072499   0.102766   0.973333                    \n",
      "    81     0.091798   0.100082   0.973333                  \n",
      "    82     0.10278    0.081863   0.973333                  \n",
      "    83     0.099819   0.089502   0.98                   \n",
      "    84     0.095093   0.088984   0.98                      \n",
      "    85     0.090794   0.074055   0.98                      \n",
      "    86     0.094213   0.071692   0.98                      \n",
      "    87     0.086386   0.080013   0.973333                  \n",
      "    88     0.073334   0.066555   0.98                      \n",
      "    89     0.079437   0.184608   0.973333                  \n",
      "    90     0.099951   0.072342   0.973333                  \n",
      "    91     0.096361   0.064167   0.98                     \n",
      "    92     0.083124   0.074168   0.973333                  \n",
      "    93     0.080067   0.075028   0.973333                  \n",
      "    94     0.069864   0.069216   0.973333                  \n",
      "    95     0.069952   0.069604   0.973333                  \n",
      "    96     0.07058    0.072871   0.973333                  \n",
      "    97     0.074311   0.067118   0.973333                  \n",
      "    98     0.05997    0.065297   0.98                      \n",
      "    99     0.062518   0.066791   0.98                      \n",
      "   100     0.068095   0.069432   0.973333                 \n",
      "   101     0.064772   0.067468   0.98                      \n",
      "   102     0.071359   0.079385   0.973333                  \n",
      "   103     0.075125   0.072833   0.973333                  \n",
      "   104     0.073732   0.068086   0.973333                  \n",
      "   105     0.064455   0.067465   0.973333                  \n",
      "   106     0.08201    0.078816   0.98                      \n",
      "   107     0.067405   0.068902   0.973333                  \n",
      "   108     0.056105   0.071219   0.973333        \n",
      "   109     0.057263   0.070736   0.973333                  \n",
      "   110     0.057808   0.067498   0.973333                  \n",
      "   111     0.049449   0.070292   0.98                      \n",
      "   112     0.050298   0.068051   0.98                      \n",
      "   113     0.05557    0.072158   0.973333                  \n",
      "   114     0.063563   0.072756   0.973333                  \n",
      "   115     0.054094   0.067397   0.98                      \n",
      "   116     0.047641   0.070014   0.98                       \n",
      "   117     0.044591   0.068302   0.98                      \n",
      "   118     0.043577   0.070267   0.973333                    \n",
      "   119     0.049635   0.070474   0.973333                  \n",
      "   120     0.046264   0.132375   0.966667                  \n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.13237517848610877, 0.9666666865348816]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr = 0.038\n",
    "learn.fit(lr, 5, cycle_len=1, cycle_mult=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.set_data(get_data(224))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "04fb158eafef42b9b585f0e8d301fe77",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=121, style=ProgressStyle(description_width='initi…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                  \n",
      "    0      0.119717   0.116435   0.973333  \n",
      "    1      0.112068   0.102661   0.966667                 \n",
      "    2      0.09827    0.084683   0.966667                  \n",
      "    3      0.079253   0.074754   0.973333                  \n",
      "    4      0.08878    0.07604    0.98                      \n",
      "    5      0.084241   0.073992   0.973333                 \n",
      "    6      0.077878   0.135541   0.966667                  \n",
      "    7      0.08287    0.078119   0.973333                  \n",
      "    8      0.083466   0.0663     0.973333                  \n",
      "    9      0.06501    0.063143   0.973333                  \n",
      "    10     0.058224   0.076933   0.973333                  \n",
      "    11     0.069879   0.073494   0.973333                  \n",
      "    12     0.066728   0.074105   0.973333                 \n",
      "    13     0.072007   0.069042   0.973333                  \n",
      "    14     0.063729   0.080178   0.973333                \n",
      "    15     0.077937   0.079291   0.98                      \n",
      "    16     0.068177   0.076044   0.98                      \n",
      "    17     0.073206   0.074471   0.98                      \n",
      "    18     0.059835   0.068185   0.98                      \n",
      "    19     0.074163   0.066942   0.973333                  \n",
      "    20     0.081743   0.072315   0.973333                  \n",
      "    21     0.096462   0.090507   0.96                      \n",
      "    22     0.079943   0.070015   0.973333                  \n",
      "    23     0.07148    0.178935   0.966667                  \n",
      "    24     0.064609   0.080454   0.966667                  \n",
      "    25     0.058807   0.174914   0.966667                  \n",
      "    26     0.055813   0.079429   0.973333                    \n",
      "    27     0.059948   0.075779   0.973333                  \n",
      "    28     0.054877   0.082209   0.973333                  \n",
      "    29     0.055255   0.089069   0.973333                  \n",
      "    30     0.048532   0.076147   0.973333                  \n",
      "    31     0.044351   0.080406   0.973333                  \n",
      "    32     0.048145   0.107353   0.96                      \n",
      "    33     0.054515   0.090961   0.966667                  \n",
      "    34     0.060815   0.083874   0.973333                  \n",
      "    35     0.063979   0.081216   0.973333                  \n",
      "    36     0.05136    0.077552   0.973333                  \n",
      "    37     0.05201    0.081281   0.973333                  \n",
      "    38     0.045751   0.079948   0.973333                  \n",
      "    39     0.04768    0.080116   0.973333                  \n",
      "    40     0.042239   0.075476   0.973333                 \n",
      "    41     0.03871    0.06757    0.973333                 \n",
      "    42     0.04344    0.101998   0.966667                  \n",
      "    43     0.103835   0.137164   0.966667                  \n",
      "    44     0.103687   0.07583    0.98                     \n",
      "    45     0.092423   0.099185   0.966667                  \n",
      "    46     0.081672   0.068929   0.973333                  \n",
      "    47     0.089026   0.088417   0.98                      \n",
      "    48     0.072366   0.072729   0.98                      \n",
      "    49     0.071682   0.092316   0.973333                  \n",
      "    50     0.066139   0.077454   0.98                      \n",
      "    51     0.05573    0.078119   0.98                      \n",
      "    52     0.076307   0.091857   0.96                      \n",
      "    53     0.068094   0.202576   0.96                      \n",
      "    54     0.092508   0.062983   0.98                      \n",
      "    55     0.085236   0.071046   0.986667                  \n",
      "    56     0.075414   0.192408   0.96                      \n",
      "    57     0.098621   0.087015   0.973333                  \n",
      "    58     0.080548   0.069323   0.966667                  \n",
      "    59     0.076158   0.106356   0.973333                    \n",
      "    60     0.077491   0.068746   0.973333                  \n",
      "    61     0.091256   0.10075    0.973333                  \n",
      "    62     0.083847   0.093728   0.973333                  \n",
      "    63     0.084123   0.072638   0.98                      \n",
      "    64     0.076005   0.070087   0.98                      \n",
      "    65     0.070216   0.173521   0.96                      \n",
      "    66     0.061489   0.071674   0.966667                  \n",
      "    67     0.059841   0.060659   0.973333                  \n",
      "    68     0.050514   0.060719   0.98                     \n",
      "    69     0.060834   0.071529   0.973333                 \n",
      "    70     0.050548   0.062842   0.973333                  \n",
      "    71     0.052303   0.06552    0.98                      \n",
      "    72     0.048422   0.073504   0.98                      \n",
      "    73     0.056725   0.068445   0.973333                  \n",
      "    74     0.048029   0.094954   0.973333                 \n",
      "    75     0.047478   0.077328   0.98                      \n",
      "    76     0.052059   0.091602   0.973333                  \n",
      "    77     0.055392   0.059057   0.973333                  \n",
      "    78     0.070878   0.081505   0.973333                  \n",
      "    79     0.071723   0.098552   0.973333                  \n",
      "    80     0.058483   0.069725   0.98                      \n",
      "    81     0.05338    0.067733   0.98                      \n",
      "    82     0.067388   0.068738   0.973333                  \n",
      "    83     0.056782   0.068744   0.98                      \n",
      "    84     0.049561   0.071497   0.98                      \n",
      "    85     0.049443   0.081114   0.973333                  \n",
      "    86     0.043749   0.065862   0.98                        \n",
      "    87     0.053255   0.061142   0.98                      \n",
      "    88     0.057496   0.06466    0.98                      \n",
      "    89     0.054002   0.073815   0.98                        \n",
      "    90     0.075948   0.066262   0.98                      \n",
      "    91     0.065409   0.061496   0.98                      \n",
      "    92     0.056036   0.066964   0.98                      \n",
      "    93     0.059089   0.079842   0.98                      \n",
      "    94     0.045769   0.07551    0.98                      \n",
      "    95     0.040842   0.061854   0.98                        \n",
      "    96     0.046401   0.064637   0.98                      \n",
      "    97     0.068453   0.067516   0.98                     \n",
      "    98     0.056283   0.067932   0.98                      \n",
      "    99     0.047608   0.067909   0.98                      \n",
      "   100     0.048727   0.070499   0.98                      \n",
      "   101     0.049271   0.0728     0.98                      \n",
      "   102     0.041121   0.071167   0.98                      \n",
      "   103     0.034277   0.073716   0.98                      \n",
      "   104     0.03656    0.073367   0.98                     \n",
      "   105     0.033626   0.075498   0.98                      \n",
      "   106     0.045774   0.074986   0.98                      \n",
      "   107     0.05367    0.07262    0.98                        \n",
      "   108     0.04478    0.069876   0.98                      \n",
      "   109     0.044719   0.070871   0.98                      \n",
      "   110     0.048021   0.071301   0.98                      \n",
      "   111     0.04694    0.071657   0.98                      \n",
      "   112     0.042818   0.071477   0.98                     \n",
      "   113     0.046756   0.07162    0.98                      \n",
      "   114     0.041692   0.070291   0.98                      \n",
      "   115     0.040318   0.133118   0.973333                  \n",
      "   116     0.045947   0.068519   0.98                      \n",
      "   117     0.047395   0.06819    0.98                      \n",
      "   118     0.047163   0.072441   0.98                      \n",
      "   119     0.046368   0.069806   0.98                        \n",
      "   120     0.044843   0.109116   0.973333                  \n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.10911554875783622, 0.9733333587646484]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.fit(lr, 5, cycle_len=1, cycle_mult=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                     \r"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.98"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_preds,y = learn.TTA()\n",
    "probs = np.mean(np.exp(log_preds),0)\n",
    "accuracy_np(probs, y)"
   ]
  }
 ],
 "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
}