Experiments.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 450,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "import os.path\n",
    "\n",
    "import keras\n",
    "import sklearn\n",
    "import seaborn\n",
    "import pandas\n",
    "import numpy\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "from microesc import urbansound8k, preprocess, train, test, report, stm32convert, stats, settings, models, features, common\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>slice_file_name</th>\n",
       "      <th>fsID</th>\n",
       "      <th>start</th>\n",
       "      <th>end</th>\n",
       "      <th>salience</th>\n",
       "      <th>fold</th>\n",
       "      <th>classID</th>\n",
       "      <th>class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>100032-3-0-0.wav</td>\n",
       "      <td>100032</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.317551</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>3</td>\n",
       "      <td>dog_bark</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>100263-2-0-117.wav</td>\n",
       "      <td>100263</td>\n",
       "      <td>58.5</td>\n",
       "      <td>62.500000</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>children_playing</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>100263-2-0-121.wav</td>\n",
       "      <td>100263</td>\n",
       "      <td>60.5</td>\n",
       "      <td>64.500000</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>children_playing</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>100263-2-0-126.wav</td>\n",
       "      <td>100263</td>\n",
       "      <td>63.0</td>\n",
       "      <td>67.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>children_playing</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>100263-2-0-137.wav</td>\n",
       "      <td>100263</td>\n",
       "      <td>68.5</td>\n",
       "      <td>72.500000</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>children_playing</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      slice_file_name    fsID  start        end  salience  fold  classID  \\\n",
       "0    100032-3-0-0.wav  100032    0.0   0.317551         1     5        3   \n",
       "1  100263-2-0-117.wav  100263   58.5  62.500000         1     5        2   \n",
       "2  100263-2-0-121.wav  100263   60.5  64.500000         1     5        2   \n",
       "3  100263-2-0-126.wav  100263   63.0  67.000000         1     5        2   \n",
       "4  100263-2-0-137.wav  100263   68.5  72.500000         1     5        2   \n",
       "\n",
       "              class  \n",
       "0          dog_bark  \n",
       "1  children_playing  \n",
       "2  children_playing  \n",
       "3  children_playing  \n",
       "4  children_playing  "
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = urbansound8k.load_dataset()\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fac12b85240>"
      ]
     },
     "execution_count": 157,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "(data.end - data.start).hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 454,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>epoch</th>\n",
       "      <th>acc</th>\n",
       "      <th>loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>voted_val_acc</th>\n",
       "      <th>fold</th>\n",
       "      <th>experiment</th>\n",
       "      <th>run</th>\n",
       "      <th>model_path</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.158667</td>\n",
       "      <td>2.440752</td>\n",
       "      <td>0.261731</td>\n",
       "      <td>2.244147</td>\n",
       "      <td>0.285859</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold3/e01-v2....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.206000</td>\n",
       "      <td>2.280682</td>\n",
       "      <td>0.298654</td>\n",
       "      <td>2.167407</td>\n",
       "      <td>0.327273</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold3/e02-v2....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.228367</td>\n",
       "      <td>2.231905</td>\n",
       "      <td>0.271154</td>\n",
       "      <td>2.143402</td>\n",
       "      <td>0.301010</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold3/e03-v2....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.241067</td>\n",
       "      <td>2.194359</td>\n",
       "      <td>0.321731</td>\n",
       "      <td>2.081261</td>\n",
       "      <td>0.329293</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold3/e04-v2....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>0.254700</td>\n",
       "      <td>2.162270</td>\n",
       "      <td>0.316923</td>\n",
       "      <td>2.081928</td>\n",
       "      <td>0.342424</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold3/e05-v2....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch       acc      loss   val_acc  val_loss  voted_val_acc  fold  \\\n",
       "0      1  0.158667  2.440752  0.261731  2.244147       0.285859     3   \n",
       "1      2  0.206000  2.280682  0.298654  2.167407       0.327273     3   \n",
       "2      3  0.228367  2.231905  0.271154  2.143402       0.301010     3   \n",
       "3      4  0.241067  2.194359  0.321731  2.081261       0.329293     3   \n",
       "4      5  0.254700  2.162270  0.316923  2.081928       0.342424     3   \n",
       "\n",
       "  experiment                 run  \\\n",
       "0          8  20190501-0223-bd5a   \n",
       "1          8  20190501-0223-bd5a   \n",
       "2          8  20190501-0223-bd5a   \n",
       "3          8  20190501-0223-bd5a   \n",
       "4          8  20190501-0223-bd5a   \n",
       "\n",
       "                                          model_path  \n",
       "0  data/models/8-20190501-0223-bd5a-fold3/e01-v2....  \n",
       "1  data/models/8-20190501-0223-bd5a-fold3/e02-v2....  \n",
       "2  data/models/8-20190501-0223-bd5a-fold3/e03-v2....  \n",
       "3  data/models/8-20190501-0223-bd5a-fold3/e04-v2....  \n",
       "4  data/models/8-20190501-0223-bd5a-fold3/e05-v2....  "
      ]
     },
     "execution_count": 454,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hist = test.load_train_history('data/models/', '20190501-0223-bd5a')\n",
    "hist.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 475,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>epoch</th>\n",
       "      <th>acc</th>\n",
       "      <th>loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>voted_val_acc</th>\n",
       "      <th>fold</th>\n",
       "      <th>experiment</th>\n",
       "      <th>run</th>\n",
       "      <th>model_path</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>experiment</th>\n",
       "      <th>fold</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"9\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <th>19</th>\n",
       "      <td>20</td>\n",
       "      <td>0.418133</td>\n",
       "      <td>1.661511</td>\n",
       "      <td>0.638846</td>\n",
       "      <td>1.297131</td>\n",
       "      <td>0.715922</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold0/e20-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>94</th>\n",
       "      <td>95</td>\n",
       "      <td>0.491367</td>\n",
       "      <td>1.355219</td>\n",
       "      <td>0.585962</td>\n",
       "      <td>1.347271</td>\n",
       "      <td>0.644144</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold1/e95-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>93</th>\n",
       "      <td>94</td>\n",
       "      <td>0.495933</td>\n",
       "      <td>1.339999</td>\n",
       "      <td>0.570385</td>\n",
       "      <td>1.452481</td>\n",
       "      <td>0.622703</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold2/e94-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <th>37</th>\n",
       "      <td>38</td>\n",
       "      <td>0.455233</td>\n",
       "      <td>1.501325</td>\n",
       "      <td>0.603654</td>\n",
       "      <td>1.350104</td>\n",
       "      <td>0.683838</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold3/e38-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <th>90</th>\n",
       "      <td>91</td>\n",
       "      <td>0.485900</td>\n",
       "      <td>1.358728</td>\n",
       "      <td>0.726731</td>\n",
       "      <td>0.993305</td>\n",
       "      <td>0.795940</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold4/e91-v0....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <th>92</th>\n",
       "      <td>93</td>\n",
       "      <td>0.494767</td>\n",
       "      <td>1.339303</td>\n",
       "      <td>0.625577</td>\n",
       "      <td>1.282158</td>\n",
       "      <td>0.704739</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold5/e93-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <th>65</th>\n",
       "      <td>66</td>\n",
       "      <td>0.483267</td>\n",
       "      <td>1.396453</td>\n",
       "      <td>0.620962</td>\n",
       "      <td>1.213497</td>\n",
       "      <td>0.702864</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold6/e66-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>52</th>\n",
       "      <td>53</td>\n",
       "      <td>0.482800</td>\n",
       "      <td>1.430011</td>\n",
       "      <td>0.623654</td>\n",
       "      <td>1.296959</td>\n",
       "      <td>0.693548</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold7/e53-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>80</th>\n",
       "      <td>81</td>\n",
       "      <td>0.486600</td>\n",
       "      <td>1.375160</td>\n",
       "      <td>0.691154</td>\n",
       "      <td>1.172646</td>\n",
       "      <td>0.737745</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/0-20190501-0223-bd5a-fold8/e81-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"9\" valign=\"top\">1</th>\n",
       "      <th>0</th>\n",
       "      <th>27</th>\n",
       "      <td>28</td>\n",
       "      <td>0.400233</td>\n",
       "      <td>1.727395</td>\n",
       "      <td>0.663269</td>\n",
       "      <td>1.354498</td>\n",
       "      <td>0.733104</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold0/e28-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>80</th>\n",
       "      <td>81</td>\n",
       "      <td>0.450667</td>\n",
       "      <td>1.506510</td>\n",
       "      <td>0.587308</td>\n",
       "      <td>1.367194</td>\n",
       "      <td>0.655405</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold1/e81-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>88</th>\n",
       "      <td>89</td>\n",
       "      <td>0.474733</td>\n",
       "      <td>1.435202</td>\n",
       "      <td>0.543654</td>\n",
       "      <td>1.420430</td>\n",
       "      <td>0.593514</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold2/e89-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <th>64</th>\n",
       "      <td>65</td>\n",
       "      <td>0.444800</td>\n",
       "      <td>1.538580</td>\n",
       "      <td>0.584808</td>\n",
       "      <td>1.381121</td>\n",
       "      <td>0.643434</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold3/e65-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <th>81</th>\n",
       "      <td>82</td>\n",
       "      <td>0.458667</td>\n",
       "      <td>1.476300</td>\n",
       "      <td>0.699615</td>\n",
       "      <td>1.085299</td>\n",
       "      <td>0.751068</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold4/e82-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>0.464033</td>\n",
       "      <td>1.456582</td>\n",
       "      <td>0.657692</td>\n",
       "      <td>1.243524</td>\n",
       "      <td>0.719320</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold5/e100-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <th>88</th>\n",
       "      <td>89</td>\n",
       "      <td>0.457633</td>\n",
       "      <td>1.480643</td>\n",
       "      <td>0.616731</td>\n",
       "      <td>1.215881</td>\n",
       "      <td>0.681384</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold6/e89-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>0.473667</td>\n",
       "      <td>1.435576</td>\n",
       "      <td>0.633462</td>\n",
       "      <td>1.167727</td>\n",
       "      <td>0.698511</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold7/e100-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>82</th>\n",
       "      <td>83</td>\n",
       "      <td>0.461533</td>\n",
       "      <td>1.477567</td>\n",
       "      <td>0.658462</td>\n",
       "      <td>1.222978</td>\n",
       "      <td>0.708333</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/1-20190501-0223-bd5a-fold8/e83-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"9\" valign=\"top\">10</th>\n",
       "      <th>0</th>\n",
       "      <th>53</th>\n",
       "      <td>54</td>\n",
       "      <td>0.376167</td>\n",
       "      <td>1.767524</td>\n",
       "      <td>0.599615</td>\n",
       "      <td>1.401350</td>\n",
       "      <td>0.680412</td>\n",
       "      <td>0</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold0/e54-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>87</th>\n",
       "      <td>88</td>\n",
       "      <td>0.400767</td>\n",
       "      <td>1.679639</td>\n",
       "      <td>0.489038</td>\n",
       "      <td>1.542464</td>\n",
       "      <td>0.568694</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold1/e88-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>93</th>\n",
       "      <td>94</td>\n",
       "      <td>0.411100</td>\n",
       "      <td>1.647057</td>\n",
       "      <td>0.477692</td>\n",
       "      <td>1.569756</td>\n",
       "      <td>0.523243</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold2/e94-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <th>94</th>\n",
       "      <td>95</td>\n",
       "      <td>0.414200</td>\n",
       "      <td>1.643704</td>\n",
       "      <td>0.508462</td>\n",
       "      <td>1.598826</td>\n",
       "      <td>0.594949</td>\n",
       "      <td>3</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold3/e95-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <th>84</th>\n",
       "      <td>85</td>\n",
       "      <td>0.408633</td>\n",
       "      <td>1.672860</td>\n",
       "      <td>0.645769</td>\n",
       "      <td>1.244991</td>\n",
       "      <td>0.700855</td>\n",
       "      <td>4</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold4/e85-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <th>98</th>\n",
       "      <td>99</td>\n",
       "      <td>0.413567</td>\n",
       "      <td>1.642098</td>\n",
       "      <td>0.545192</td>\n",
       "      <td>1.455317</td>\n",
       "      <td>0.608748</td>\n",
       "      <td>5</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold5/e99-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <th>80</th>\n",
       "      <td>81</td>\n",
       "      <td>0.403900</td>\n",
       "      <td>1.672058</td>\n",
       "      <td>0.570577</td>\n",
       "      <td>1.433581</td>\n",
       "      <td>0.650358</td>\n",
       "      <td>6</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold6/e81-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>94</th>\n",
       "      <td>95</td>\n",
       "      <td>0.423100</td>\n",
       "      <td>1.624424</td>\n",
       "      <td>0.594231</td>\n",
       "      <td>1.337946</td>\n",
       "      <td>0.638958</td>\n",
       "      <td>7</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold7/e95-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>98</th>\n",
       "      <td>99</td>\n",
       "      <td>0.402133</td>\n",
       "      <td>1.677952</td>\n",
       "      <td>0.614615</td>\n",
       "      <td>1.344095</td>\n",
       "      <td>0.669118</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/10-20190501-0223-bd5a-fold8/e99-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">2</th>\n",
       "      <th>0</th>\n",
       "      <th>55</th>\n",
       "      <td>56</td>\n",
       "      <td>0.423333</td>\n",
       "      <td>1.628961</td>\n",
       "      <td>0.620577</td>\n",
       "      <td>1.348940</td>\n",
       "      <td>0.681558</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/2-20190501-0223-bd5a-fold0/e56-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>83</th>\n",
       "      <td>84</td>\n",
       "      <td>0.433900</td>\n",
       "      <td>1.567785</td>\n",
       "      <td>0.571923</td>\n",
       "      <td>1.395273</td>\n",
       "      <td>0.621622</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/2-20190501-0223-bd5a-fold1/e84-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>58</th>\n",
       "      <td>59</td>\n",
       "      <td>0.420067</td>\n",
       "      <td>1.631019</td>\n",
       "      <td>0.495769</td>\n",
       "      <td>1.506978</td>\n",
       "      <td>0.551351</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/2-20190501-0223-bd5a-fold2/e59-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">6</th>\n",
       "      <th>6</th>\n",
       "      <th>53</th>\n",
       "      <td>54</td>\n",
       "      <td>0.407700</td>\n",
       "      <td>1.660430</td>\n",
       "      <td>0.580577</td>\n",
       "      <td>1.394251</td>\n",
       "      <td>0.656325</td>\n",
       "      <td>6</td>\n",
       "      <td>6</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/6-20190501-0223-bd5a-fold6/e54-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>68</th>\n",
       "      <td>69</td>\n",
       "      <td>0.420533</td>\n",
       "      <td>1.624861</td>\n",
       "      <td>0.644231</td>\n",
       "      <td>1.283020</td>\n",
       "      <td>0.686104</td>\n",
       "      <td>7</td>\n",
       "      <td>6</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/6-20190501-0223-bd5a-fold7/e69-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>0.438833</td>\n",
       "      <td>1.555874</td>\n",
       "      <td>0.645192</td>\n",
       "      <td>1.226741</td>\n",
       "      <td>0.694853</td>\n",
       "      <td>8</td>\n",
       "      <td>6</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/6-20190501-0223-bd5a-fold8/e100-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"9\" valign=\"top\">7</th>\n",
       "      <th>0</th>\n",
       "      <th>76</th>\n",
       "      <td>77</td>\n",
       "      <td>0.412767</td>\n",
       "      <td>1.643197</td>\n",
       "      <td>0.685192</td>\n",
       "      <td>1.212289</td>\n",
       "      <td>0.746850</td>\n",
       "      <td>0</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold0/e77-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>85</th>\n",
       "      <td>86</td>\n",
       "      <td>0.422733</td>\n",
       "      <td>1.612098</td>\n",
       "      <td>0.557692</td>\n",
       "      <td>1.414758</td>\n",
       "      <td>0.613739</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold1/e86-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>70</th>\n",
       "      <td>71</td>\n",
       "      <td>0.412967</td>\n",
       "      <td>1.647787</td>\n",
       "      <td>0.517885</td>\n",
       "      <td>1.456887</td>\n",
       "      <td>0.572973</td>\n",
       "      <td>2</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold2/e71-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <th>80</th>\n",
       "      <td>81</td>\n",
       "      <td>0.420000</td>\n",
       "      <td>1.613708</td>\n",
       "      <td>0.547115</td>\n",
       "      <td>1.502580</td>\n",
       "      <td>0.608081</td>\n",
       "      <td>3</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold3/e81-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <th>98</th>\n",
       "      <td>99</td>\n",
       "      <td>0.419800</td>\n",
       "      <td>1.599477</td>\n",
       "      <td>0.677692</td>\n",
       "      <td>1.144226</td>\n",
       "      <td>0.714744</td>\n",
       "      <td>4</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold4/e99-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>0.430633</td>\n",
       "      <td>1.575744</td>\n",
       "      <td>0.607308</td>\n",
       "      <td>1.382958</td>\n",
       "      <td>0.658566</td>\n",
       "      <td>5</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold5/e100-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <th>64</th>\n",
       "      <td>65</td>\n",
       "      <td>0.402067</td>\n",
       "      <td>1.673478</td>\n",
       "      <td>0.580385</td>\n",
       "      <td>1.433197</td>\n",
       "      <td>0.650358</td>\n",
       "      <td>6</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold6/e65-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>55</th>\n",
       "      <td>56</td>\n",
       "      <td>0.401600</td>\n",
       "      <td>1.694506</td>\n",
       "      <td>0.638654</td>\n",
       "      <td>1.309767</td>\n",
       "      <td>0.692308</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold7/e56-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>0.426900</td>\n",
       "      <td>1.592769</td>\n",
       "      <td>0.631731</td>\n",
       "      <td>1.250298</td>\n",
       "      <td>0.675245</td>\n",
       "      <td>8</td>\n",
       "      <td>7</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/7-20190501-0223-bd5a-fold8/e100-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"9\" valign=\"top\">8</th>\n",
       "      <th>0</th>\n",
       "      <th>82</th>\n",
       "      <td>83</td>\n",
       "      <td>0.397200</td>\n",
       "      <td>1.690526</td>\n",
       "      <td>0.640769</td>\n",
       "      <td>1.334831</td>\n",
       "      <td>0.706758</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold0/e83-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>84</th>\n",
       "      <td>85</td>\n",
       "      <td>0.408733</td>\n",
       "      <td>1.644153</td>\n",
       "      <td>0.565385</td>\n",
       "      <td>1.436173</td>\n",
       "      <td>0.635135</td>\n",
       "      <td>1</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold1/e85-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>92</th>\n",
       "      <td>93</td>\n",
       "      <td>0.416833</td>\n",
       "      <td>1.623681</td>\n",
       "      <td>0.520769</td>\n",
       "      <td>1.456549</td>\n",
       "      <td>0.557838</td>\n",
       "      <td>2</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold2/e93-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <th>96</th>\n",
       "      <td>97</td>\n",
       "      <td>0.410833</td>\n",
       "      <td>1.633083</td>\n",
       "      <td>0.568077</td>\n",
       "      <td>1.463433</td>\n",
       "      <td>0.625253</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold3/e97-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <th>84</th>\n",
       "      <td>85</td>\n",
       "      <td>0.403667</td>\n",
       "      <td>1.676084</td>\n",
       "      <td>0.660000</td>\n",
       "      <td>1.254582</td>\n",
       "      <td>0.712607</td>\n",
       "      <td>4</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold4/e85-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <th>78</th>\n",
       "      <td>79</td>\n",
       "      <td>0.398600</td>\n",
       "      <td>1.675545</td>\n",
       "      <td>0.608846</td>\n",
       "      <td>1.397934</td>\n",
       "      <td>0.668287</td>\n",
       "      <td>5</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold5/e79-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <th>64</th>\n",
       "      <td>65</td>\n",
       "      <td>0.401700</td>\n",
       "      <td>1.695186</td>\n",
       "      <td>0.576154</td>\n",
       "      <td>1.450187</td>\n",
       "      <td>0.652745</td>\n",
       "      <td>6</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold6/e65-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>81</th>\n",
       "      <td>82</td>\n",
       "      <td>0.413333</td>\n",
       "      <td>1.661191</td>\n",
       "      <td>0.628462</td>\n",
       "      <td>1.309633</td>\n",
       "      <td>0.689826</td>\n",
       "      <td>7</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold7/e82-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>92</th>\n",
       "      <td>93</td>\n",
       "      <td>0.402500</td>\n",
       "      <td>1.674773</td>\n",
       "      <td>0.618462</td>\n",
       "      <td>1.319640</td>\n",
       "      <td>0.671569</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/8-20190501-0223-bd5a-fold8/e93-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"9\" valign=\"top\">9</th>\n",
       "      <th>0</th>\n",
       "      <th>25</th>\n",
       "      <td>26</td>\n",
       "      <td>0.396467</td>\n",
       "      <td>1.717115</td>\n",
       "      <td>0.643269</td>\n",
       "      <td>1.313971</td>\n",
       "      <td>0.706758</td>\n",
       "      <td>0</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold0/e26-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <th>71</th>\n",
       "      <td>72</td>\n",
       "      <td>0.442333</td>\n",
       "      <td>1.542495</td>\n",
       "      <td>0.571154</td>\n",
       "      <td>1.398561</td>\n",
       "      <td>0.632883</td>\n",
       "      <td>1</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold1/e72-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <th>24</th>\n",
       "      <td>25</td>\n",
       "      <td>0.404267</td>\n",
       "      <td>1.707469</td>\n",
       "      <td>0.543269</td>\n",
       "      <td>1.476043</td>\n",
       "      <td>0.585946</td>\n",
       "      <td>2</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold2/e25-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <th>91</th>\n",
       "      <td>92</td>\n",
       "      <td>0.450033</td>\n",
       "      <td>1.506307</td>\n",
       "      <td>0.605577</td>\n",
       "      <td>1.307722</td>\n",
       "      <td>0.660606</td>\n",
       "      <td>3</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold3/e92-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <th>72</th>\n",
       "      <td>73</td>\n",
       "      <td>0.443067</td>\n",
       "      <td>1.540186</td>\n",
       "      <td>0.699615</td>\n",
       "      <td>1.109777</td>\n",
       "      <td>0.757479</td>\n",
       "      <td>4</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold4/e73-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <th>83</th>\n",
       "      <td>84</td>\n",
       "      <td>0.441200</td>\n",
       "      <td>1.523608</td>\n",
       "      <td>0.638654</td>\n",
       "      <td>1.277470</td>\n",
       "      <td>0.709599</td>\n",
       "      <td>5</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold5/e84-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>0.456000</td>\n",
       "      <td>1.494909</td>\n",
       "      <td>0.651923</td>\n",
       "      <td>1.196810</td>\n",
       "      <td>0.721957</td>\n",
       "      <td>6</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold6/e100-v1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <th>65</th>\n",
       "      <td>66</td>\n",
       "      <td>0.433933</td>\n",
       "      <td>1.580532</td>\n",
       "      <td>0.641538</td>\n",
       "      <td>1.253051</td>\n",
       "      <td>0.705955</td>\n",
       "      <td>7</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold7/e66-v1....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <th>85</th>\n",
       "      <td>86</td>\n",
       "      <td>0.448667</td>\n",
       "      <td>1.521484</td>\n",
       "      <td>0.654615</td>\n",
       "      <td>1.291977</td>\n",
       "      <td>0.710784</td>\n",
       "      <td>8</td>\n",
       "      <td>9</td>\n",
       "      <td>20190501-0223-bd5a</td>\n",
       "      <td>data/models/9-20190501-0223-bd5a-fold8/e86-v1....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>99 rows × 10 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    epoch       acc      loss   val_acc  val_loss  \\\n",
       "experiment fold                                                     \n",
       "0          0    19     20  0.418133  1.661511  0.638846  1.297131   \n",
       "           1    94     95  0.491367  1.355219  0.585962  1.347271   \n",
       "           2    93     94  0.495933  1.339999  0.570385  1.452481   \n",
       "           3    37     38  0.455233  1.501325  0.603654  1.350104   \n",
       "           4    90     91  0.485900  1.358728  0.726731  0.993305   \n",
       "           5    92     93  0.494767  1.339303  0.625577  1.282158   \n",
       "           6    65     66  0.483267  1.396453  0.620962  1.213497   \n",
       "           7    52     53  0.482800  1.430011  0.623654  1.296959   \n",
       "           8    80     81  0.486600  1.375160  0.691154  1.172646   \n",
       "1          0    27     28  0.400233  1.727395  0.663269  1.354498   \n",
       "           1    80     81  0.450667  1.506510  0.587308  1.367194   \n",
       "           2    88     89  0.474733  1.435202  0.543654  1.420430   \n",
       "           3    64     65  0.444800  1.538580  0.584808  1.381121   \n",
       "           4    81     82  0.458667  1.476300  0.699615  1.085299   \n",
       "           5    99    100  0.464033  1.456582  0.657692  1.243524   \n",
       "           6    88     89  0.457633  1.480643  0.616731  1.215881   \n",
       "           7    99    100  0.473667  1.435576  0.633462  1.167727   \n",
       "           8    82     83  0.461533  1.477567  0.658462  1.222978   \n",
       "10         0    53     54  0.376167  1.767524  0.599615  1.401350   \n",
       "           1    87     88  0.400767  1.679639  0.489038  1.542464   \n",
       "           2    93     94  0.411100  1.647057  0.477692  1.569756   \n",
       "           3    94     95  0.414200  1.643704  0.508462  1.598826   \n",
       "           4    84     85  0.408633  1.672860  0.645769  1.244991   \n",
       "           5    98     99  0.413567  1.642098  0.545192  1.455317   \n",
       "           6    80     81  0.403900  1.672058  0.570577  1.433581   \n",
       "           7    94     95  0.423100  1.624424  0.594231  1.337946   \n",
       "           8    98     99  0.402133  1.677952  0.614615  1.344095   \n",
       "2          0    55     56  0.423333  1.628961  0.620577  1.348940   \n",
       "           1    83     84  0.433900  1.567785  0.571923  1.395273   \n",
       "           2    58     59  0.420067  1.631019  0.495769  1.506978   \n",
       "...                   ...       ...       ...       ...       ...   \n",
       "6          6    53     54  0.407700  1.660430  0.580577  1.394251   \n",
       "           7    68     69  0.420533  1.624861  0.644231  1.283020   \n",
       "           8    99    100  0.438833  1.555874  0.645192  1.226741   \n",
       "7          0    76     77  0.412767  1.643197  0.685192  1.212289   \n",
       "           1    85     86  0.422733  1.612098  0.557692  1.414758   \n",
       "           2    70     71  0.412967  1.647787  0.517885  1.456887   \n",
       "           3    80     81  0.420000  1.613708  0.547115  1.502580   \n",
       "           4    98     99  0.419800  1.599477  0.677692  1.144226   \n",
       "           5    99    100  0.430633  1.575744  0.607308  1.382958   \n",
       "           6    64     65  0.402067  1.673478  0.580385  1.433197   \n",
       "           7    55     56  0.401600  1.694506  0.638654  1.309767   \n",
       "           8    99    100  0.426900  1.592769  0.631731  1.250298   \n",
       "8          0    82     83  0.397200  1.690526  0.640769  1.334831   \n",
       "           1    84     85  0.408733  1.644153  0.565385  1.436173   \n",
       "           2    92     93  0.416833  1.623681  0.520769  1.456549   \n",
       "           3    96     97  0.410833  1.633083  0.568077  1.463433   \n",
       "           4    84     85  0.403667  1.676084  0.660000  1.254582   \n",
       "           5    78     79  0.398600  1.675545  0.608846  1.397934   \n",
       "           6    64     65  0.401700  1.695186  0.576154  1.450187   \n",
       "           7    81     82  0.413333  1.661191  0.628462  1.309633   \n",
       "           8    92     93  0.402500  1.674773  0.618462  1.319640   \n",
       "9          0    25     26  0.396467  1.717115  0.643269  1.313971   \n",
       "           1    71     72  0.442333  1.542495  0.571154  1.398561   \n",
       "           2    24     25  0.404267  1.707469  0.543269  1.476043   \n",
       "           3    91     92  0.450033  1.506307  0.605577  1.307722   \n",
       "           4    72     73  0.443067  1.540186  0.699615  1.109777   \n",
       "           5    83     84  0.441200  1.523608  0.638654  1.277470   \n",
       "           6    99    100  0.456000  1.494909  0.651923  1.196810   \n",
       "           7    65     66  0.433933  1.580532  0.641538  1.253051   \n",
       "           8    85     86  0.448667  1.521484  0.654615  1.291977   \n",
       "\n",
       "                    voted_val_acc  fold experiment                 run  \\\n",
       "experiment fold                                                          \n",
       "0          0    19       0.715922     0          0  20190501-0223-bd5a   \n",
       "           1    94       0.644144     1          0  20190501-0223-bd5a   \n",
       "           2    93       0.622703     2          0  20190501-0223-bd5a   \n",
       "           3    37       0.683838     3          0  20190501-0223-bd5a   \n",
       "           4    90       0.795940     4          0  20190501-0223-bd5a   \n",
       "           5    92       0.704739     5          0  20190501-0223-bd5a   \n",
       "           6    65       0.702864     6          0  20190501-0223-bd5a   \n",
       "           7    52       0.693548     7          0  20190501-0223-bd5a   \n",
       "           8    80       0.737745     8          0  20190501-0223-bd5a   \n",
       "1          0    27       0.733104     0          1  20190501-0223-bd5a   \n",
       "           1    80       0.655405     1          1  20190501-0223-bd5a   \n",
       "           2    88       0.593514     2          1  20190501-0223-bd5a   \n",
       "           3    64       0.643434     3          1  20190501-0223-bd5a   \n",
       "           4    81       0.751068     4          1  20190501-0223-bd5a   \n",
       "           5    99       0.719320     5          1  20190501-0223-bd5a   \n",
       "           6    88       0.681384     6          1  20190501-0223-bd5a   \n",
       "           7    99       0.698511     7          1  20190501-0223-bd5a   \n",
       "           8    82       0.708333     8          1  20190501-0223-bd5a   \n",
       "10         0    53       0.680412     0         10  20190501-0223-bd5a   \n",
       "           1    87       0.568694     1         10  20190501-0223-bd5a   \n",
       "           2    93       0.523243     2         10  20190501-0223-bd5a   \n",
       "           3    94       0.594949     3         10  20190501-0223-bd5a   \n",
       "           4    84       0.700855     4         10  20190501-0223-bd5a   \n",
       "           5    98       0.608748     5         10  20190501-0223-bd5a   \n",
       "           6    80       0.650358     6         10  20190501-0223-bd5a   \n",
       "           7    94       0.638958     7         10  20190501-0223-bd5a   \n",
       "           8    98       0.669118     8         10  20190501-0223-bd5a   \n",
       "2          0    55       0.681558     0          2  20190501-0223-bd5a   \n",
       "           1    83       0.621622     1          2  20190501-0223-bd5a   \n",
       "           2    58       0.551351     2          2  20190501-0223-bd5a   \n",
       "...                           ...   ...        ...                 ...   \n",
       "6          6    53       0.656325     6          6  20190501-0223-bd5a   \n",
       "           7    68       0.686104     7          6  20190501-0223-bd5a   \n",
       "           8    99       0.694853     8          6  20190501-0223-bd5a   \n",
       "7          0    76       0.746850     0          7  20190501-0223-bd5a   \n",
       "           1    85       0.613739     1          7  20190501-0223-bd5a   \n",
       "           2    70       0.572973     2          7  20190501-0223-bd5a   \n",
       "           3    80       0.608081     3          7  20190501-0223-bd5a   \n",
       "           4    98       0.714744     4          7  20190501-0223-bd5a   \n",
       "           5    99       0.658566     5          7  20190501-0223-bd5a   \n",
       "           6    64       0.650358     6          7  20190501-0223-bd5a   \n",
       "           7    55       0.692308     7          7  20190501-0223-bd5a   \n",
       "           8    99       0.675245     8          7  20190501-0223-bd5a   \n",
       "8          0    82       0.706758     0          8  20190501-0223-bd5a   \n",
       "           1    84       0.635135     1          8  20190501-0223-bd5a   \n",
       "           2    92       0.557838     2          8  20190501-0223-bd5a   \n",
       "           3    96       0.625253     3          8  20190501-0223-bd5a   \n",
       "           4    84       0.712607     4          8  20190501-0223-bd5a   \n",
       "           5    78       0.668287     5          8  20190501-0223-bd5a   \n",
       "           6    64       0.652745     6          8  20190501-0223-bd5a   \n",
       "           7    81       0.689826     7          8  20190501-0223-bd5a   \n",
       "           8    92       0.671569     8          8  20190501-0223-bd5a   \n",
       "9          0    25       0.706758     0          9  20190501-0223-bd5a   \n",
       "           1    71       0.632883     1          9  20190501-0223-bd5a   \n",
       "           2    24       0.585946     2          9  20190501-0223-bd5a   \n",
       "           3    91       0.660606     3          9  20190501-0223-bd5a   \n",
       "           4    72       0.757479     4          9  20190501-0223-bd5a   \n",
       "           5    83       0.709599     5          9  20190501-0223-bd5a   \n",
       "           6    99       0.721957     6          9  20190501-0223-bd5a   \n",
       "           7    65       0.705955     7          9  20190501-0223-bd5a   \n",
       "           8    85       0.710784     8          9  20190501-0223-bd5a   \n",
       "\n",
       "                                                           model_path  \n",
       "experiment fold                                                        \n",
       "0          0    19  data/models/0-20190501-0223-bd5a-fold0/e20-v1....  \n",
       "           1    94  data/models/0-20190501-0223-bd5a-fold1/e95-v1....  \n",
       "           2    93  data/models/0-20190501-0223-bd5a-fold2/e94-v1....  \n",
       "           3    37  data/models/0-20190501-0223-bd5a-fold3/e38-v1....  \n",
       "           4    90  data/models/0-20190501-0223-bd5a-fold4/e91-v0....  \n",
       "           5    92  data/models/0-20190501-0223-bd5a-fold5/e93-v1....  \n",
       "           6    65  data/models/0-20190501-0223-bd5a-fold6/e66-v1....  \n",
       "           7    52  data/models/0-20190501-0223-bd5a-fold7/e53-v1....  \n",
       "           8    80  data/models/0-20190501-0223-bd5a-fold8/e81-v1....  \n",
       "1          0    27  data/models/1-20190501-0223-bd5a-fold0/e28-v1....  \n",
       "           1    80  data/models/1-20190501-0223-bd5a-fold1/e81-v1....  \n",
       "           2    88  data/models/1-20190501-0223-bd5a-fold2/e89-v1....  \n",
       "           3    64  data/models/1-20190501-0223-bd5a-fold3/e65-v1....  \n",
       "           4    81  data/models/1-20190501-0223-bd5a-fold4/e82-v1....  \n",
       "           5    99  data/models/1-20190501-0223-bd5a-fold5/e100-v1...  \n",
       "           6    88  data/models/1-20190501-0223-bd5a-fold6/e89-v1....  \n",
       "           7    99  data/models/1-20190501-0223-bd5a-fold7/e100-v1...  \n",
       "           8    82  data/models/1-20190501-0223-bd5a-fold8/e83-v1....  \n",
       "10         0    53  data/models/10-20190501-0223-bd5a-fold0/e54-v1...  \n",
       "           1    87  data/models/10-20190501-0223-bd5a-fold1/e88-v1...  \n",
       "           2    93  data/models/10-20190501-0223-bd5a-fold2/e94-v1...  \n",
       "           3    94  data/models/10-20190501-0223-bd5a-fold3/e95-v1...  \n",
       "           4    84  data/models/10-20190501-0223-bd5a-fold4/e85-v1...  \n",
       "           5    98  data/models/10-20190501-0223-bd5a-fold5/e99-v1...  \n",
       "           6    80  data/models/10-20190501-0223-bd5a-fold6/e81-v1...  \n",
       "           7    94  data/models/10-20190501-0223-bd5a-fold7/e95-v1...  \n",
       "           8    98  data/models/10-20190501-0223-bd5a-fold8/e99-v1...  \n",
       "2          0    55  data/models/2-20190501-0223-bd5a-fold0/e56-v1....  \n",
       "           1    83  data/models/2-20190501-0223-bd5a-fold1/e84-v1....  \n",
       "           2    58  data/models/2-20190501-0223-bd5a-fold2/e59-v1....  \n",
       "...                                                               ...  \n",
       "6          6    53  data/models/6-20190501-0223-bd5a-fold6/e54-v1....  \n",
       "           7    68  data/models/6-20190501-0223-bd5a-fold7/e69-v1....  \n",
       "           8    99  data/models/6-20190501-0223-bd5a-fold8/e100-v1...  \n",
       "7          0    76  data/models/7-20190501-0223-bd5a-fold0/e77-v1....  \n",
       "           1    85  data/models/7-20190501-0223-bd5a-fold1/e86-v1....  \n",
       "           2    70  data/models/7-20190501-0223-bd5a-fold2/e71-v1....  \n",
       "           3    80  data/models/7-20190501-0223-bd5a-fold3/e81-v1....  \n",
       "           4    98  data/models/7-20190501-0223-bd5a-fold4/e99-v1....  \n",
       "           5    99  data/models/7-20190501-0223-bd5a-fold5/e100-v1...  \n",
       "           6    64  data/models/7-20190501-0223-bd5a-fold6/e65-v1....  \n",
       "           7    55  data/models/7-20190501-0223-bd5a-fold7/e56-v1....  \n",
       "           8    99  data/models/7-20190501-0223-bd5a-fold8/e100-v1...  \n",
       "8          0    82  data/models/8-20190501-0223-bd5a-fold0/e83-v1....  \n",
       "           1    84  data/models/8-20190501-0223-bd5a-fold1/e85-v1....  \n",
       "           2    92  data/models/8-20190501-0223-bd5a-fold2/e93-v1....  \n",
       "           3    96  data/models/8-20190501-0223-bd5a-fold3/e97-v1....  \n",
       "           4    84  data/models/8-20190501-0223-bd5a-fold4/e85-v1....  \n",
       "           5    78  data/models/8-20190501-0223-bd5a-fold5/e79-v1....  \n",
       "           6    64  data/models/8-20190501-0223-bd5a-fold6/e65-v1....  \n",
       "           7    81  data/models/8-20190501-0223-bd5a-fold7/e82-v1....  \n",
       "           8    92  data/models/8-20190501-0223-bd5a-fold8/e93-v1....  \n",
       "9          0    25  data/models/9-20190501-0223-bd5a-fold0/e26-v1....  \n",
       "           1    71  data/models/9-20190501-0223-bd5a-fold1/e72-v1....  \n",
       "           2    24  data/models/9-20190501-0223-bd5a-fold2/e25-v1....  \n",
       "           3    91  data/models/9-20190501-0223-bd5a-fold3/e92-v1....  \n",
       "           4    72  data/models/9-20190501-0223-bd5a-fold4/e73-v1....  \n",
       "           5    83  data/models/9-20190501-0223-bd5a-fold5/e84-v1....  \n",
       "           6    99  data/models/9-20190501-0223-bd5a-fold6/e100-v1...  \n",
       "           7    65  data/models/9-20190501-0223-bd5a-fold7/e66-v1....  \n",
       "           8    85  data/models/9-20190501-0223-bd5a-fold8/e86-v1....  \n",
       "\n",
       "[99 rows x 10 columns]"
      ]
     },
     "execution_count": 475,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# best.nickname == 'Stride-DS-5x5'\n",
    "best = test.pick_best(hist, n_best=1)\n",
    "best"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 486,
   "metadata": {},
   "outputs": [],
   "source": [
    "best = best.reset_index(level=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 487,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "fold\n",
       "0    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "1    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "2    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "3    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "4    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "5    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "6    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "7    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "8    AxesSubplot(0.125,0.125;0.775x0.755)\n",
       "Name: voted_val_acc, dtype: object"
      ]
     },
     "execution_count": 487,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "best.groupby(level='fold').voted_val_acc.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 461,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'data/models/3-20190501-0223-bd5a-fold8/e79-v1.25.t1.59.model.hdf5'"
      ]
     },
     "execution_count": 461,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_path = best.query('fold == \"8\" and experiment == \"3\"').model_path[0]\n",
    "model_path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1653, 9)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>slice_file_name</th>\n",
       "      <th>fsID</th>\n",
       "      <th>start</th>\n",
       "      <th>end</th>\n",
       "      <th>salience</th>\n",
       "      <th>fold</th>\n",
       "      <th>classID</th>\n",
       "      <th>class</th>\n",
       "      <th>set</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>100648-1-0-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>4.823402</td>\n",
       "      <td>5.471927</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>100648-1-1-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>8.998279</td>\n",
       "      <td>10.052132</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>67</th>\n",
       "      <td>101729-0-0-1.wav</td>\n",
       "      <td>101729</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>2</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>air_conditioner</td>\n",
       "      <td>val</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68</th>\n",
       "      <td>101729-0-0-11.wav</td>\n",
       "      <td>101729</td>\n",
       "      <td>5.500000</td>\n",
       "      <td>9.500000</td>\n",
       "      <td>2</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>air_conditioner</td>\n",
       "      <td>val</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      slice_file_name    fsID     start        end  salience  fold  classID  \\\n",
       "9    100648-1-0-0.wav  100648  4.823402   5.471927         2    10        1   \n",
       "10   100648-1-1-0.wav  100648  8.998279  10.052132         2    10        1   \n",
       "67   101729-0-0-1.wav  101729  0.500000   4.500000         2     9        0   \n",
       "68  101729-0-0-11.wav  101729  5.500000   9.500000         2     9        0   \n",
       "\n",
       "              class   set  \n",
       "9          car_horn  test  \n",
       "10         car_horn  test  \n",
       "67  air_conditioner   val  \n",
       "68  air_conditioner   val  "
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def create_eval_set(fold):\n",
    "    data = urbansound8k.load_dataset()\n",
    "    folds, test  = urbansound8k.folds(data)\n",
    "    test = test.copy()\n",
    "    train = folds[fold][0].copy()\n",
    "    val = folds[fold][1].copy()\n",
    "\n",
    "    test['set'] = 'test'\n",
    "    train['set'] = 'train'\n",
    "    val['set'] = 'val'\n",
    "    # XXX: include train?\n",
    "    df = pandas.concat([test, val])\n",
    "    return df\n",
    "     \n",
    "eval_set = create_eval_set(fold=8)\n",
    "print(eval_set.shape)\n",
    "eval_set.groupby('set').head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Warning: Sample Sample(start=2.8792743764172335, end=3.24140589569161, fold=10, slice_file_name='26344-4-1-0.wav') with start 2.8792743764172335 has 0 length\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>slice_file_name</th>\n",
       "      <th>fsID</th>\n",
       "      <th>start</th>\n",
       "      <th>end</th>\n",
       "      <th>salience</th>\n",
       "      <th>fold</th>\n",
       "      <th>classID</th>\n",
       "      <th>class</th>\n",
       "      <th>set</th>\n",
       "      <th>predictions</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>100648-1-0-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>4.823402</td>\n",
       "      <td>5.471927</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "      <td>[0.03206492, 0.12941872, 0.11333097, 0.1137044...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>100648-1-1-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>8.998279</td>\n",
       "      <td>10.052132</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "      <td>[0.014229942, 0.16539712, 0.066213205, 0.14090...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>67</th>\n",
       "      <td>101729-0-0-1.wav</td>\n",
       "      <td>101729</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>2</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>air_conditioner</td>\n",
       "      <td>val</td>\n",
       "      <td>[0.09555469, 0.010715012, 0.03366288, 0.101037...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68</th>\n",
       "      <td>101729-0-0-11.wav</td>\n",
       "      <td>101729</td>\n",
       "      <td>5.500000</td>\n",
       "      <td>9.500000</td>\n",
       "      <td>2</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "      <td>air_conditioner</td>\n",
       "      <td>val</td>\n",
       "      <td>[0.18683255, 0.015797311, 0.06032902, 0.090838...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      slice_file_name    fsID     start        end  salience  fold  classID  \\\n",
       "9    100648-1-0-0.wav  100648  4.823402   5.471927         2    10        1   \n",
       "10   100648-1-1-0.wav  100648  8.998279  10.052132         2    10        1   \n",
       "67   101729-0-0-1.wav  101729  0.500000   4.500000         2     9        0   \n",
       "68  101729-0-0-11.wav  101729  5.500000   9.500000         2     9        0   \n",
       "\n",
       "              class   set                                        predictions  \n",
       "9          car_horn  test  [0.03206492, 0.12941872, 0.11333097, 0.1137044...  \n",
       "10         car_horn  test  [0.014229942, 0.16539712, 0.066213205, 0.14090...  \n",
       "67  air_conditioner   val  [0.09555469, 0.010715012, 0.03366288, 0.101037...  \n",
       "68  air_conditioner   val  [0.18683255, 0.015797311, 0.06032902, 0.090838...  "
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "exsettings = settings.load_settings(common.load_settings_path('experiments/ldcnn20k60.yaml'))\n",
    "\n",
    "def load_sample(sample):\n",
    "    fsettings = features.settings(exsettings)\n",
    "    return features.load_sample(sample, fsettings, start_time=sample.start,\n",
    "                window_frames=exsettings['frames'], feature_dir='data/features')\n",
    "\n",
    "def predict(model, data):\n",
    "    return features.predict_voted(exsettings, model, data, loader=load_sample,\n",
    "                                    method=exsettings['voting'], overlap=exsettings['voting_overlap'])\n",
    "\n",
    "def model_predict(predictor, model_path, data):\n",
    "    model = keras.models.load_model(model_path)\n",
    "    p = predictor(model, data)\n",
    "    return p\n",
    "\n",
    "p = model_predict(predict, model_path, eval_set)\n",
    "eval_set['predictions'] = list(p)\n",
    "eval_set.groupby('set').head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>slice_file_name</th>\n",
       "      <th>fsID</th>\n",
       "      <th>start</th>\n",
       "      <th>end</th>\n",
       "      <th>salience</th>\n",
       "      <th>fold</th>\n",
       "      <th>classID</th>\n",
       "      <th>class</th>\n",
       "      <th>set</th>\n",
       "      <th>predictions</th>\n",
       "      <th>best_p</th>\n",
       "      <th>best_y</th>\n",
       "      <th>correct</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>100648-1-0-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>4.823402</td>\n",
       "      <td>5.471927</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "      <td>[0.03206492, 0.12941872, 0.11333097, 0.1137044...</td>\n",
       "      <td>0.235730</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>100648-1-1-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>8.998279</td>\n",
       "      <td>10.052132</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "      <td>[0.014229942, 0.16539712, 0.066213205, 0.14090...</td>\n",
       "      <td>0.226467</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>100648-1-2-0.wav</td>\n",
       "      <td>100648</td>\n",
       "      <td>16.699509</td>\n",
       "      <td>17.104837</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>car_horn</td>\n",
       "      <td>test</td>\n",
       "      <td>[0.0050243563, 0.14825162, 0.009926457, 0.1720...</td>\n",
       "      <td>0.342725</td>\n",
       "      <td>4</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     slice_file_name    fsID      start        end  salience  fold  classID  \\\n",
       "9   100648-1-0-0.wav  100648   4.823402   5.471927         2    10        1   \n",
       "10  100648-1-1-0.wav  100648   8.998279  10.052132         2    10        1   \n",
       "11  100648-1-2-0.wav  100648  16.699509  17.104837         2    10        1   \n",
       "\n",
       "       class   set                                        predictions  \\\n",
       "9   car_horn  test  [0.03206492, 0.12941872, 0.11333097, 0.1137044...   \n",
       "10  car_horn  test  [0.014229942, 0.16539712, 0.066213205, 0.14090...   \n",
       "11  car_horn  test  [0.0050243563, 0.14825162, 0.009926457, 0.1720...   \n",
       "\n",
       "      best_p  best_y  correct  \n",
       "9   0.235730       4    False  \n",
       "10  0.226467       4    False  \n",
       "11  0.342725       4    False  "
      ]
     },
     "execution_count": 212,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_set['class'] = eval_set.classID.apply(lambda i: urbansound8k.classnames[i])\n",
    "eval_set['best_p'] = eval_set.predictions.apply(lambda p: numpy.max(p))\n",
    "eval_set['best_y'] = eval_set.predictions.apply(lambda p: numpy.argmax(p))\n",
    "eval_set['correct'] = eval_set.best_y == eval_set.classID\n",
    "eval_set.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.37763323783874514"
      ]
     },
     "execution_count": 210,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def threshold(df, q=0.8):\n",
    "    q = df[df.correct == False].best_p.quantile(q=q)\n",
    "    return q\n",
    "    \n",
    "threshold(eval_set[eval_set.set == 'val'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 448,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "class\n",
       "air_conditioner     0.368953\n",
       "car_horn            0.383808\n",
       "children_playing    0.372627\n",
       "dog_bark            0.357520\n",
       "drilling            0.555217\n",
       "engine_idling       0.291193\n",
       "gun_shot            0.338796\n",
       "jackhammer          0.632930\n",
       "siren               0.259379\n",
       "street_music        0.308365\n",
       "dtype: float64"
      ]
     },
     "execution_count": 448,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_set[eval_set.set == 'val'].groupby('class').apply(threshold)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 398,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f02b68308d0>"
      ]
     },
     "execution_count": 398,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "def plot_errors(df, ax=None, q=0.8, bins=20, ylim=None):\n",
    "    \n",
    "    if ax is None:\n",
    "        fig, ax = plt.subplots(1)\n",
    "    \n",
    "    (_, wrong), (isright, right) = df.groupby('correct')\n",
    "    assert isright == True\n",
    "    \n",
    "    # FIXME: add legend\n",
    "    right.best_p.hist(ax=ax, color='green', alpha=0.4, bins=bins)\n",
    "    wrong.best_p.hist(ax=ax, color='red', alpha=0.4, bins=bins)\n",
    "    \n",
    "    ax.set_xlim((0, 1.0))\n",
    "    if ylim:\n",
    "        ax.set_ylim(ylim)\n",
    "    \n",
    "    # Illustrate a threshold\n",
    "    if q is not None:\n",
    "        p_min = threshold(df, q=q)\n",
    "        ax.axvline(x=p_min, color='black', linewidth=1.0, alpha=0.7)\n",
    "\n",
    "    return ax\n",
    "    \n",
    "\n",
    "plot_errors(eval_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 407,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_errors_classwise(df, figsize=(12,4)):\n",
    "\n",
    "    groups = eval_set.groupby('class')\n",
    "    fig, axs = plt.subplots(2, len(groups)//2, figsize=figsize)\n",
    "    \n",
    "    for i, (classname, data) in enumerate(groups):\n",
    "        x = i // 2\n",
    "        y = i % 2\n",
    "        ax = axs[y, x]\n",
    "        ax.set_title(classname)\n",
    "        if x != 0:\n",
    "            ax.tick_params(labelleft=False)\n",
    "        if y == 0:\n",
    "            ax.tick_params(labelbottom=False)\n",
    "        bins = numpy.linspace(0, 1.0, 20)\n",
    "        plot_errors(data, ax=ax, ylim=(0, 30), q=None, bins=bins)\n",
    "    \n",
    "    fig.tight_layout()\n",
    "    return fig\n",
    "\n",
    "plot_errors_classwise(eval_set[eval_set.set == 'val']);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 447,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def score(df, average=None, threshold=0.0):\n",
    "    y_true = df.classID\n",
    "    y_pred = df.best_y\n",
    "\n",
    "    uncertain = df.best_p < threshold\n",
    "    uncertain_ratio = numpy.count_nonzero(uncertain.astype(int)) / len(y_pred)\n",
    "    y_pred = y_pred.mask(uncertain, 11) # unknown class\n",
    "  \n",
    "    labels = list(range(0,10)) # don't count unknown class\n",
    "    precision = sklearn.metrics.precision_score(y_true, y_pred, average=average, labels=labels)\n",
    "    recall = sklearn.metrics.recall_score(y_true, y_pred, average=average, labels=labels)\n",
    "    out = pandas.Series({\n",
    "        'precision': precision,\n",
    "        'recall': recall,    \n",
    "        'uncertain': uncertain_ratio,\n",
    "    })\n",
    "    return out\n",
    "\n",
    "def plot_precision_recall(data, ax=None):\n",
    "    df = pandas.DataFrame({\n",
    "        'threshold': numpy.linspace(0, 1.0, 50, endpoint=False)\n",
    "    })\n",
    "\n",
    "    micro = df.apply(lambda r: score(data, average='micro', threshold=r.threshold), axis=1)\n",
    "    micro['threshold'] = df.threshold\n",
    "    micro['micro'] = micro.precision\n",
    "    \n",
    "    macro = df.apply(lambda r: score(data, average='macro', threshold=r.threshold), axis=1)\n",
    "    macro['threshold'] = df.threshold\n",
    "    macro['macro'] = macro.precision\n",
    "\n",
    "    \n",
    "    fig, (ax2, ax) = plt.subplots(1, 2, figsize=(8, 4))\n",
    "    \n",
    "    macro.plot.line(ax=ax2, y=['precision', 'recall'], x='threshold', ylim=(0.0, 1.0))\n",
    "    \n",
    "    micro.plot.line(ax=ax, y='micro', x='recall', ylim=(0.0, 1.0), xlim=(0, 1))\n",
    "    macro.plot.line(ax=ax, y='macro', x='recall', ylim=(0.0, 1.0), xlim=(0, 1))\n",
    "\n",
    "\n",
    "    ax.set_yticks(numpy.arange(0.0, 1.0, 0.1))\n",
    "    ax.grid(True)\n",
    "    ax.set_ylabel('Precision')\n",
    "    ax.set_xlabel('Recall')\n",
    "    ax.set_aspect('equal')\n",
    "    \n",
    "    ax2.set_aspect('equal')\n",
    "    ax2.set_yticks(numpy.arange(0.0, 1.0, 0.1))\n",
    "    ax2.grid(True)\n",
    "    ax2.set_xlabel('Probability threshold for \"unknown\" class')\n",
    "    ax2.set_ylabel('Performance metric')\n",
    "    \n",
    "    fig.tight_layout()\n",
    "\n",
    "plot_precision_recall(eval_set[eval_set.set == 'val'])\n",
    "plot_precision_recall(eval_set[eval_set.set == 'test'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cpu_mhz</th>\n",
       "      <th>cycles_avg</th>\n",
       "      <th>cycles_macc</th>\n",
       "      <th>duration_avg</th>\n",
       "      <th>macc</th>\n",
       "      <th>stack</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>experiment</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>80</td>\n",
       "      <td>77743690</td>\n",
       "      <td>7.632552</td>\n",
       "      <td>0.971796</td>\n",
       "      <td>10185806</td>\n",
       "      <td>388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>80</td>\n",
       "      <td>5858230</td>\n",
       "      <td>13.551088</td>\n",
       "      <td>0.073227</td>\n",
       "      <td>432307</td>\n",
       "      <td>292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>80</td>\n",
       "      <td>6527943</td>\n",
       "      <td>13.678647</td>\n",
       "      <td>0.081599</td>\n",
       "      <td>477236</td>\n",
       "      <td>292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>80</td>\n",
       "      <td>4932455</td>\n",
       "      <td>14.508458</td>\n",
       "      <td>0.061655</td>\n",
       "      <td>339971</td>\n",
       "      <td>292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>80</td>\n",
       "      <td>26011364</td>\n",
       "      <td>8.726309</td>\n",
       "      <td>0.325142</td>\n",
       "      <td>2980798</td>\n",
       "      <td>276</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            cpu_mhz  cycles_avg  cycles_macc  duration_avg      macc  stack\n",
       "experiment                                                                 \n",
       "0                80    77743690     7.632552      0.971796  10185806    388\n",
       "4                80     5858230    13.551088      0.073227    432307    292\n",
       "2                80     6527943    13.678647      0.081599    477236    292\n",
       "3                80     4932455    14.508458      0.061655    339971    292\n",
       "1                80    26011364     8.726309      0.325142   2980798    276"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import json\n",
    "\n",
    "# TODO: check that MACC is correct wrt given by stm32conv\n",
    "def load_device_results(results_dir):\n",
    "\n",
    "    frames = []\n",
    "    \n",
    "    for filename in os.listdir(results_dir):\n",
    "        if filename.endswith('.device.json'):\n",
    "            experiment = filename.rstrip('.device.json')\n",
    "            p = os.path.join(results_dir, filename)\n",
    "            with open(p, 'r') as f:\n",
    "                contents = f.read()\n",
    "                contents = contents.replace(\"'\", '\"') # hack invalid JSON\n",
    "                d = json.loads(contents)\n",
    "                d['experiment'] = experiment\n",
    "                df = pandas.DataFrame([d])\n",
    "                frames.append(df)\n",
    "\n",
    "    df = pandas.concat(frames)\n",
    "    df.set_index('experiment', inplace=True)\n",
    "    return df\n",
    "\n",
    "run = '20190412-1024'\n",
    "results_dir = os.path.join('./data/results', run)\n",
    "df = load_device_results(results_dir)\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f231703b940>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "df.plot.scatter(y='cycles_avg', x='duration_avg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f23170cacc0>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "df.plot.scatter(y='cycles_macc', x='macc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'maccs_frame': 843838,\n",
       " 'flash_usage': 287460.0,\n",
       " 'ram_usage_max': 56250.0,\n",
       " 'ram_usage_min': 56250.0}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "s = {\n",
    "    'model': 'strided',\n",
    "    'frames': 31,\n",
    "    'n_mels': 60,\n",
    "    'samplerate': 22050,\n",
    "    'conv_size': (3,3),\n",
    "    'conv_block': 'effnet',\n",
    "    'downsample_size': (2,2),\n",
    "    'n_stages': 3,\n",
    "    'n_blocks_per_stage': 2,\n",
    "    'filters': 20,\n",
    "}\n",
    "model = models.build(s)\n",
    "ss, ll = stats.check_model_constraints(model)\n",
    "ss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 60, 31, 1)         0         \n",
      "_________________________________________________________________\n",
      "conv0 (Conv2D)               (None, 30, 16, 20)        200       \n",
      "_________________________________________________________________\n",
      "conv0_bn (BatchNormalization (None, 30, 16, 20)        80        \n",
      "_________________________________________________________________\n",
      "conv0_relu (Activation)      (None, 30, 16, 20)        0         \n",
      "_________________________________________________________________\n",
      "conv1pw (Conv2D)             (None, 30, 16, 10)        200       \n",
      "_________________________________________________________________\n",
      "conv1pw_bn (BatchNormalizati (None, 30, 16, 10)        40        \n",
      "_________________________________________________________________\n",
      "conv1pw_relu (Activation)    (None, 30, 16, 10)        0         \n",
      "_________________________________________________________________\n",
      "conv1dwv (DepthwiseConv2D)   (None, 30, 16, 10)        30        \n",
      "_________________________________________________________________\n",
      "conv1dwv_bn (BatchNormalizat (None, 30, 16, 10)        40        \n",
      "_________________________________________________________________\n",
      "conv1dwv_relu (Activation)   (None, 30, 16, 10)        0         \n",
      "_________________________________________________________________\n",
      "conv1dwh (DepthwiseConv2D)   (None, 30, 16, 10)        30        \n",
      "_________________________________________________________________\n",
      "conv1dwh_bn (BatchNormalizat (None, 30, 16, 10)        40        \n",
      "_________________________________________________________________\n",
      "conv1dwh_relu (Activation)   (None, 30, 16, 10)        0         \n",
      "_________________________________________________________________\n",
      "conv1rh (Conv2D)             (None, 30, 16, 20)        200       \n",
      "_________________________________________________________________\n",
      "conv1rh_bn (BatchNormalizati (None, 30, 16, 20)        80        \n",
      "_________________________________________________________________\n",
      "conv1rh_relu (Activation)    (None, 30, 16, 20)        0         \n",
      "_________________________________________________________________\n",
      "conv2pw (Conv2D)             (None, 15, 8, 17)         340       \n",
      "_________________________________________________________________\n",
      "conv2pw_bn (BatchNormalizati (None, 15, 8, 17)         68        \n",
      "_________________________________________________________________\n",
      "conv2pw_relu (Activation)    (None, 15, 8, 17)         0         \n",
      "_________________________________________________________________\n",
      "conv2dwv (DepthwiseConv2D)   (None, 15, 8, 17)         51        \n",
      "_________________________________________________________________\n",
      "conv2dwv_bn (BatchNormalizat (None, 15, 8, 17)         68        \n",
      "_________________________________________________________________\n",
      "conv2dwv_relu (Activation)   (None, 15, 8, 17)         0         \n",
      "_________________________________________________________________\n",
      "conv2dwh (DepthwiseConv2D)   (None, 15, 8, 17)         51        \n",
      "_________________________________________________________________\n",
      "conv2dwh_bn (BatchNormalizat (None, 15, 8, 17)         68        \n",
      "_________________________________________________________________\n",
      "conv2dwh_relu (Activation)   (None, 15, 8, 17)         0         \n",
      "_________________________________________________________________\n",
      "conv2rh (Conv2D)             (None, 15, 8, 35)         595       \n",
      "_________________________________________________________________\n",
      "conv2rh_bn (BatchNormalizati (None, 15, 8, 35)         140       \n",
      "_________________________________________________________________\n",
      "conv2rh_relu (Activation)    (None, 15, 8, 35)         0         \n",
      "_________________________________________________________________\n",
      "conv3pw (Conv2D)             (None, 15, 8, 17)         595       \n",
      "_________________________________________________________________\n",
      "conv3pw_bn (BatchNormalizati (None, 15, 8, 17)         68        \n",
      "_________________________________________________________________\n",
      "conv3pw_relu (Activation)    (None, 15, 8, 17)         0         \n",
      "_________________________________________________________________\n",
      "conv3dwv (DepthwiseConv2D)   (None, 15, 8, 17)         51        \n",
      "_________________________________________________________________\n",
      "conv3dwv_bn (BatchNormalizat (None, 15, 8, 17)         68        \n",
      "_________________________________________________________________\n",
      "conv3dwv_relu (Activation)   (None, 15, 8, 17)         0         \n",
      "_________________________________________________________________\n",
      "conv3dwh (DepthwiseConv2D)   (None, 15, 8, 17)         51        \n",
      "_________________________________________________________________\n",
      "conv3dwh_bn (BatchNormalizat (None, 15, 8, 17)         68        \n",
      "_________________________________________________________________\n",
      "conv3dwh_relu (Activation)   (None, 15, 8, 17)         0         \n",
      "_________________________________________________________________\n",
      "conv3rh (Conv2D)             (None, 15, 8, 35)         595       \n",
      "_________________________________________________________________\n",
      "conv3rh_bn (BatchNormalizati (None, 15, 8, 35)         140       \n",
      "_________________________________________________________________\n",
      "conv3rh_relu (Activation)    (None, 15, 8, 35)         0         \n",
      "_________________________________________________________________\n",
      "conv4pw (Conv2D)             (None, 8, 4, 30)          1050      \n",
      "_________________________________________________________________\n",
      "conv4pw_bn (BatchNormalizati (None, 8, 4, 30)          120       \n",
      "_________________________________________________________________\n",
      "conv4pw_relu (Activation)    (None, 8, 4, 30)          0         \n",
      "_________________________________________________________________\n",
      "conv4dwv (DepthwiseConv2D)   (None, 8, 4, 30)          90        \n",
      "_________________________________________________________________\n",
      "conv4dwv_bn (BatchNormalizat (None, 8, 4, 30)          120       \n",
      "_________________________________________________________________\n",
      "conv4dwv_relu (Activation)   (None, 8, 4, 30)          0         \n",
      "_________________________________________________________________\n",
      "conv4dwh (DepthwiseConv2D)   (None, 8, 4, 30)          90        \n",
      "_________________________________________________________________\n",
      "conv4dwh_bn (BatchNormalizat (None, 8, 4, 30)          120       \n",
      "_________________________________________________________________\n",
      "conv4dwh_relu (Activation)   (None, 8, 4, 30)          0         \n",
      "_________________________________________________________________\n",
      "conv4rh (Conv2D)             (None, 8, 4, 61)          1830      \n",
      "_________________________________________________________________\n",
      "conv4rh_bn (BatchNormalizati (None, 8, 4, 61)          244       \n",
      "_________________________________________________________________\n",
      "conv4rh_relu (Activation)    (None, 8, 4, 61)          0         \n",
      "_________________________________________________________________\n",
      "conv5pw (Conv2D)             (None, 8, 4, 30)          1830      \n",
      "_________________________________________________________________\n",
      "conv5pw_bn (BatchNormalizati (None, 8, 4, 30)          120       \n",
      "_________________________________________________________________\n",
      "conv5pw_relu (Activation)    (None, 8, 4, 30)          0         \n",
      "_________________________________________________________________\n",
      "conv5dwv (DepthwiseConv2D)   (None, 8, 2, 30)          90        \n",
      "_________________________________________________________________\n",
      "conv5dwv_bn (BatchNormalizat (None, 8, 2, 30)          120       \n",
      "_________________________________________________________________\n",
      "conv5dwv_relu (Activation)   (None, 8, 2, 30)          0         \n",
      "_________________________________________________________________\n",
      "conv5dwh (DepthwiseConv2D)   (None, 8, 2, 30)          90        \n",
      "_________________________________________________________________\n",
      "conv5dwh_bn (BatchNormalizat (None, 8, 2, 30)          120       \n",
      "_________________________________________________________________\n",
      "conv5dwh_relu (Activation)   (None, 8, 2, 30)          0         \n",
      "_________________________________________________________________\n",
      "conv5rh (Conv2D)             (None, 8, 2, 61)          1830      \n",
      "_________________________________________________________________\n",
      "conv5rh_bn (BatchNormalizati (None, 8, 2, 61)          244       \n",
      "_________________________________________________________________\n",
      "conv5rh_relu (Activation)    (None, 8, 2, 61)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 976)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 64)                62528     \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                650       \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 10)                0         \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 10)                0         \n",
      "=================================================================\n",
      "Total params: 75,243\n",
      "Trainable params: 74,155\n",
      "Non-trainable params: 1,088\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUIAAAFpCAYAAADpxiFpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3XmcXGWZ9//PNyExJEEjIThoNMGJGQwgWxYgixHZcWMbBBmN4iM4BsaZl/Mo8gDR+aHDPKOOiqIoKgMMhFVQVIYBkzRLQhYii0hIpBGEhxBhGIIEJX39/jh3dYqmurs6qapzTtf3/Xr1q6vOcp+7qsPFfZbruhURmJm1syF5d8DMLG8OhGbW9hwIzaztORCaWdtzIDSztudAaGZtz4HQzNqeA6GZtT0HQjNrew6EZtb2tsu7A+1gzI5jYuc37Zx3N8wGjaEayqjho/rdbuXKlRsiYlx/2zkQtsDOb9qZK26+Iu9umA0aT//xaQ6fdHi/20l6tJ72fGpsZm2vtCNCSXdGxIENbnMicGBE/Ecf2xwC/DMwHPgT8I8RcVtf7T6yYSPv+ebtDezpq/309FlNbd9sMCvtiLDRQTCZCJzUzzYbgPdGxJ7AR4BLm9APM2uh0gZCSRvT77mSFkm6RtJvJF0uSWldp6TzJd2dfial5T+SdFzPtshGerMlrZb097WOGxH3RMQT6e0DwAhJr2nW5zSz5ittIOxhH+DTwBTgrcDMqnX/ExHTgQuAf+unnc8BHRGxd0R8rY7jHgvcExEv9Vwh6ROSVkha0fXHP9b1IcwsH4MlEN4dEY9HRBewmuwUt+KKqt8HNOqAknYHzgdOrbU+Ii6KiKkRMXXIyJGNOqyZNcFgCYTVI7LNvPImUNR4/TLps6fT6OEDOZik8cD1wIcjYt2Ae2tmhVLau8YDcALZtb8TgLvSsk5gP+Aq4P3AsLT8eWCHvhqTNAa4CTgzIu6opwO77jSaK3xX16ywBsuIsC+vkbQM+DugcgPke8A7Jd0NzABeSMvvBV6W9KvebpYA84FJwNnppspqSU4bMSsxDeZZ7CR1AlMjYkOe/Zi85+RwZolZ4wwgs2RlREztb7t2GBGamfVpUF8jjIiJW7uvpMPI7gpXeyQijt6mTplZ4QzqQLgtIuJm4Oa8+2FmzedAOECSzgROIXtM54wUMPvUilxjKy7ngRefA+EASJoCfBDYHXgj8F+SJkfE5nx7ZmbbYlDeLJH0YUn3psdgLpU0QdKtadmtkt6StvuRpG9IulPSbyv5x5IWSjqyqr0fSTqW7JnDKyPipYh4BFgLTM/jM5pZ4wy6QJhS384CDoqIvcieH7wA+PeIeAdwOfCNql12AWYB7yF78BrgSrIHsJE0HHg38DPgTcBjVfs+npbV6odzjc1KYtAFQuAg4JrKs4MR8QxZjnGlxuClZIGv4scR0RURvwbekJb9HDgoVZU5AlgSES8CqnG8mg9iOtfYrDwGYyAUvQSnKtXrq/OUBRARm4BFwGFkI8Mr0/rHgTdXbT8eeAIzK7XBGAhvBf5a0lgASTsCd5Ld5AD4EFDPLdwrgY8Cs9nyGM2NwAclvUbSrsDbgLsb2Hczy8Ggu2scEQ9IOg9YLGkzcA9wBvADSf8IPE0W4Przn8C/AzdGxJ+q2r4K+DVZBZtP1XPH2EUXzIpt0AVCgIi4BLikx+KDamw3r8f70VWv/wyMrbHPecB5DemomRXCYDw1NjMbEAdCM2t7DoRm1vbaMhBKGivpl5I2Srog7/6YWb4G5c2SOmwCzgb2SD9N1e5FF1x0wIqu0CPCZuUMR8QLEXE7WUDsecyNkr4iaVU6xjhJO0tamdbvJSmqjr1OklNHzEqssIGwyTnDfRkFrIqIfYHFwLkRsZ5sIvfXkj1gvYJsIvgJwPqIcDKxWYkVNhDS3JzhvnQBC9Pry6qOcSfZxPFzgC+l37OBjlqNuOiCWXkUORA2M2d4ICrH6CALfBOAG4C9yILkkpo7ueiCWWkUORA2M2e4L0OA49Lrk6qOsQQ4GXg4IrqAZ4AjgbrmNjaz4irsXeNm5gxD91SfrwWGS/oAcGg6rX4B2D3dHHmOdI0xIjolwZYR4O3A+Ih4tr8OONfYrNgG9bzGW0PSxuqc40bwvMZmjeV5jc3MGsyBsIdGjwbNrPgcCM2s7TkQmlnbK+xd40ZIj95cA0wDfhQR83vZbh4wtbf128q5xr5jbsU2qAMhLS6uYGblVIhT45yKK3xU0hpJi8lS55A0NLUrSWMkdUmak9Z1SJosqVPSmKp21kp6Q8/2zaw8cg+EeRRXkLQL8AWyAHgIMAUgTcS0Jr2fBawkK67wGrKHp9eQpdcdndqZAXRGxFPb9i2YWZ5yD4TkU1xhBrAoIp5O2SYLq9Z1kBVUmAN8OR17GrA8rV9ICrpk6X7V+3Zz0QWz8ihCIMyruEJvx6wUV5hONqocA8xlS2rdXcAkSeOADwDX1WzcRRfMSqMIgTCP4grLgLmpZP8w4Pge6w4EulKAXQ2cSiq3FVlO4vXAV4EHI+IPdfTNzAos97vGeRVXkLSAbHT3JLAKGJr685Kkx4ClqYkO4ETgvqpjLSQ7VZ5Xz2d00QWzYnPRhRZw0QWzxnLRBTOzBnMgNLO250BoZm1vUAfCeidylzSv3oneJS2Q9JnG9dLM8pb7XeMmK0SusYsu+I65FVshRoQlyjWelHafImlR2vaMJn89ZtZkuQfCkuUar01N7EaWxTIdODc9lG1mJZV7IKR8ucYAN0XES6nP66v60c25xmblUYRAWLZc45592EyNa63ONTYrjyIEwlLlGpvZ4JP7XeOS5hoPiHONzYrNucYt4Fxjs8ZyrrGZWYM5EJpZ23MgNLO250BoZm0v97vGRVDvRPBp206yyeA31Nt+u+catzvnWhefA2GmruIMkoa2rEdm1jKlOjXOqTjDRklflLSMLPUP4HRJqyTdJ2m3Fnx0M2ui0gTCPIozJKOA+yNiRgqWABsiYl/gQsC1Cc1KrjSBkHyKM0CWS3xtj2WVuYxXAhNr7eSiC2blUaZAmFdxhk2pPFe1Sts1Cy6kY7nogllJlCkQ5lGcwczaQGnuGudVnKERfXfRBbNiK00gBIiIS4BLeiw+qMZ283q8H131+s/A2Br7TOzlmKN72y4iVpDVKjSzEivTqbGZWVM4EJpZ23MgNLO215aBUNIhklamzJCVkl51ndHM2kepbpY00AbgvRHxhKQ9yB6jeVOzDtbKogtO8DcbuEKPCJuYW3xPRDyRFj8AjEjZJpXc4q+kXOJbJY2TtLOklWn9XpKi6tjrJPmJabMSK2wgbGFu8bHAPRFRyRYZBaxKucSLgXMjYj1ZsHwt2YPYK8gmfp8ArI8I59CZlVhhAyEtyC1OwfZ8suk6K7rYMuH7ZVXHuBOYSTbx+5fS79n0Ms2nc43NyqPIgbCpucWSxgPXAx+OiHV1HKMy8fsE4AZgL7IguaTmTs41NiuNIgfCpuUWSxoD3AScGRF39Nh+CHBcen1S1TGWACcDD0dEF/AMcCTQc38zK5nC3jVucm7xfGAScLaks9OyQ9O1wBeA3dPNkedI1xgjolMSbBkB3g6Mj4hn++uAc43Nis0TvPcgaWPP/OJt5QnezRrLE7ybmTWYA2EPjR4NmlnxORCaWdtzIDSztjeoA2G9xRUkzZN0Qav7Z2bFUNjHZxqkpcUVeuOiC2bFVogRYU7FFT4qaY2kxWSpc0gamtqVpDGSuiTNSes6JE2W1JkeyK4ca62kSkqfmZVQ7oEwj+IKknYBvkAWAA8BpgCkaTvXpPezyOYtnp2C5/iIWEOWXnd0OtYMoDMinmrMt2Fmecg9EJJPcYUZwKKIeDplmyzc0jwdZAUV5gBfTseeBixP6xeSgi5Zul/1vt1cdMGsPIoQCPMqrtDbMSvFFaaTjSrHkM1UV0mtuwuYJGkc8AHgupoddtEFs9IoQiDMo7jCMmCupLGShgHH91h3INCVAuxqspFkB0BkOYnXA18FHoyIPwz4E5tZoeR+1zin4gpPSlpANrp7ElgFDE39eUnSY8DStH0HcCJwX9WxFpKdKs+r5zO66IJZsbnoQgu46IJZY7nogplZgzkQmlnbcyA0s7Y3qANhM3KNJS2Q9JnG9tTM8pT7XeMmc66xmfWrECPCEuUaT0ptTZG0KG17Rmu+JTNrltwDYclyjdemtnYjy2KZDpybHso2s5LKPRBSvlxjgJsi4qXU5/VV/ejmXGOz8ihCICxbrnHPPmymxrVW5xqblUcRAmGpco3NbPDJ/a5xSXONB8S5xmbF5lzjFnCusVljOdfYzKzBHAjNrO05EJpZ23MgNLO250BI/cUZ0radknZqZf/MrLlyf3ymIOoqziBp6NY03sqiC3lwoQcru1KNCHMqzrBR0hclLSNL/QM4XdKqNILcrXXfgJk1Q2kCYR7FGdL7UcD9ETEjIirDug0RsS9wIeDahGYlV5pASD7FGSDLJb62R18qcxmvBCbW6qyLLpiVR5kCYV7FGTal8lzVKm3XLLiQjuWiC2YlUaZAmEdxBjNrA6W5a5xTcYb1jei7iy6YFZuLLrSAiy6YNZaLLpiZNZgDoZm1PQdCM2t7bRkIJU2XtDr9/ErS0Xn3yczyU5q7xg12PzA1Il5OU3v+StJPIuLlZhzME7ybFVuhR4RNzC3+Y1XQG0HVg9gpt/grKZf4VknjJO0saWVav5ekqDr2Okl+YtqsxAobCJudWyxphqQHyCZlOq0qMI4CVqVc4sXAuel5whGSXkv2IPYKsonfJwDrI8I5dGYlVthASJNziyNiWUTsTjZ5+5mSRqR9utgy4ftlVce4E5hJNvH7l9Lv2fQyzadzjc3Ko8iBsKm5xd0NRDwIvADs0c8xKhO/TwBuAPYiC5JLau7kXGOz0ihyIGxmbvGukrZLrycAfwV0pu2HAMel1ydVHWMJcDLwcER0Ac8ARwLOTTYrucLeNW5ybvEs4HOS/kx2Kvy3lVNwstHh7unmyHOka4wR0SkJtowAbwfGR8Sz/XXAucZmxeZc4x4kbYyI0Y1s07nGZo3lXGMzswZzIOyh0aNBMys+B0Iza3sOhGbW9gZ1IKy3uIKkeZIuaHX/zKwYCvv4TIO0tLhCb1x0wazYCjEizKm4wkclrZG0mCx1DklDU7uSNEZSl6Q5aV2HpMmSOtNkT5V21kqqpPSZWQnlHgjzKK6QRodfIAuAhwBTANK0nWvS+1lk8xbPTrnK4yNiDVl63dGVtoHOiHiq0d+LmbVO7oGQfIorzAAWRcTTKdtkYVX7HWQFFeYAX07HngYsT+sXkoIuWbpf9b7dXHTBrDyKEAjzKq7Q2zErxRWmk40qxwBz2ZJadxcwSdI44APAdTU77KILZqVRhECYR3GFZcBcSWMlDQOOr2pnGXAg0JUC7GrgVFK5rchyEq8Hvgo8GBF/2LqPbWZFkftd47yKK0haQDa6exJYBQxN/XlJ0mPA0tRGB3Ai2TXGioVkp8rz6vmMLrpgVmwuutACLrpg1lguumBm1mAOhGbW9uoKhJJGSjpb0vfS+7dJek9zu2Zm1hr1jgh/SPbYygHp/ePA/9eUHpmZtVi9d43/MiJOkHQiQES8qFS3vsgkTQcuqrwFFkTE9TW2m0eWkzy/jjYXABsj4l/r7Ydzjc2Krd5A+CdJ25MeQpb0l7zyweaiKkTRBTMrtnpPjc8FfgG8WdLlZA9B/+9GdaJERRcmpd2nSFqUtj2jUd+DmeWjrkAYEbcAx5A9QHwF2ShrUSM6ULKiC2vT8XYjS+ebDpybslN6fi7nGpuVxEAen3kTWfbFcGCOpGMa1IeyFV0AuCkiXkp9Xl/Vj27ONTYrj7quEUr6AfAO4AGyVDXITjNrFhwYoIYUXZC0iC1FF16VxhERD0qqt+jCacAbgXOAf+SVRRd69mEzBUhVNLOtV++IcP80uvlIRHw0/XysQX0oVdEFMxt86h3J3CVpSjodbaiSFl0YEBddMCu2uooupDunPwH+H9lpocgqUr2jud0bHFx0wayxGl10od4R4Q+AvyEbFXX1s62ZWanUGwh/FxE3NrUnZmY5qTcQ/kbSf5CdHnffMY2IRtw1NjPLVb2BcHuyAHho1bJGPT6Tu3pzktO2nWQPlG9oUffMrMnqCoQRUc9d2zKrKydZ0tCtabyVRRfy4EIPVnb11iMcIelTkr4t6QeVn2Z3rkY/8shJ3ijpi5KWsaUM2emSVkm6T9Jurfn0ZtYs9T5QfSnwF2SZG4uB8cDzzepULXnkJKd9RgH3R8SMiKgM6zZExL7AhcBnmvF5zax16g2EkyLibOCFiLgEOArYs3ndqimPnGTIUuiu7dGXyrXRlcDEWp110QWz8qg3EP45/f5vSXsAr6OXANBEeU0EvylVpalWabvXPGMXXTArj3oD4UWSXg/8H+BG4NfA+U3rVW155CSbWRuo9/GZS4FjyUaBl6Rlryo91Ux55SQ3gnONzYqt3kB4A/Ac2TWx3Er0p+uTl/RYfFCN7eb1eD+66vWfgbE91l9KFuxrHXN0j/cTq16vICvRZWYlVm8gHB8R/Wc4m5mVUL3XCO+U1Oq7xGZmLVHviHAWME/SI7gMl5kNMvUGwiOa2gszsxzVm2v8aLM7koeUkvdrsiILdU/YPlBbk2vs/F2z1hnILHaD0dfIsk3MrI0VOhA2q8hCev0B4LdkM/NVH3OjpK+kogq3ShonaWdJK9P6vSRF1bHXSXLqiFmJFTYQNrPIgqRRwGfJJnnvaRSwKhVVWAycGxHrgRGSXkuWkbKCbOL3CcD6iHAysVmJFTYQ0twiC18AvhYRG2sct4stE75fVnWMO4GZZBO/fyn9nk0v03y66IJZeRR5YvJmTvw+AzhO0r8AY4AuSZsi4oI+jtFBFvgmkGXafDat+2nNjkVcRKp6PWyXN/Y/VaCZ5abII8KmFVmIiNkRMTGly/0b8KWqIDgEOC69PqnqGEuAk4GHI6ILeAY4Erhjaz+gmRVDYUeETS6y0JcXgN3TzZHnSNcYI6JTEmQBEbIAOT4inu2vQRddMCu2uiZ4byeSNvYstLCtPMG7WWM1eoL3Ip8am5m1hANhD40eDZpZ8TkQmlnba4tAKOktKWOk5oxzkuZJqvXojJm1gcLeNW6wXHOKXXTBrNgKMSLMKaf4o5LWSFpMljGCpKGpXUkaI6lL0py0rkPSZEmdksZUtbNWUkvnbzGzxso9EOaRUyxpl7RsJnAIMAUgTdu5Jr2fRTZHy+yUojc+ItaQZZUcndqZAXRGxFON+j7MrPVyD4Tkk1M8A1gUEU+nh6wXVq3rIMsjngN8OR17GrA8rV9ICrpkWS7V+3ZzrrFZeRQhEDZz4vYZwL9I6gQ+DXxe0vwabVar5BRPB35Glos8ly0ZJXcBkySNAz4AXFezw57g3aw0ihAI88gpXgbMlTRW0jDg+Kp2lgEHAl0pwK4GTiVVmYksFed64KvAgxHxh6394GZWDLnfNc4jpzginpS0gGx09ySwChia1r0k6TFgadq8AzgRuK+qiYVkp8rz6vmMzjU2KzbnGreAc43NGsu5xmZmDeZAaGZtz4HQzNqeA6GZtb3c7xq3Qn8TuUuaB0yNiPk919XYdgGwcSATwjvX2KzY2mVE6InczaxXhQiEJSq6MCntPkXSorTtGc3+fsysuXIPhCUrurA2NbEbWTrfdODclJ1iZiWVeyCkfEUXAG6KiJdSn9dX9aObiy6YlUcRbpa0fCJ34PE+jtkBnAa8ETgH+EdeWXShZx82U+N79ATvZuVRhBFhqYoumNngk/uIsKRFFwbERRfMis1FF1rARRfMGstFF8zMGsyB0MzangOhmbU9B8Iq6mci+LRNp6SdWtkvM2uu3O8aF0yfOcmShm5No1tTdGFruFCD2dYp1Ygwp5zkjZK+KGkZWcYLwOmSVkm6T9JurfjsZtY8pQmEeeQkJ6OA+yNiRkRUhnUbImJf4EKg19NoMyuH0gRC8slJhiyF7toeyypzGa8EJtbqrHONzcqjTNcIW56TnNLxNqWqNNUqbdfMM07Hcq6xWUmUaUSYR06ymbWB0owI88hJbhTnGpsVW2kCIUBEXAJc0mPxQTW2m9fj/eiq138GxvZxjAW97ZveT6x6vYKsRJeZlViZTo3NzJrCgdDM2p4DoZm1PQdCM2t7bRkIJU2U9KKk1ennO3n3yczyU6q7xg22LiL2bsWBtqboggsomLVOoUeEzSyy0McxN0r6SiqqcKukcZJ2lrQyrd9LUlQde52kkc37Fsys2QobCJtZZCGt21XSPZIWS5pd1c4oYFUqqrAYODci1gMjJL2WLCNlBdnE7xOA9RHhZGKzEitsIKS5RRaeBN4SEfsA/wD8RwpyAF1smfD9sqpj3AnMJJv4/Uvp92x6mebTRRfMyqPIgbAhRRaARWwpsnBlWv5SRPwhvV4JrAMm93OMDrLANwG4AdiLLEguqblTxEURMTUipg4Z6TNnsyIrciBsWpGFdN1vaHr9VuBtZEVZIftOjkuvT6o6xhLgZODhiOgCngGOBO7Yys9nZgVR2LvGTS6yMAf4oqSXyUppnZZOvQFeAHZPN0eeI11jjIhOSbBlBHg7MD4inu2vAy66YFZsnuC9B0kbexZa2Fae4N2ssTzBu5lZgzkQ9tDo0aCZFZ8DoZm1vUEdCOvNKZY0T5JL85u1qcLeNW6gluUU96a3XGPnE5sVQyFGhDnlFH9U0hpJi8kyRpA0NLUrSWMkdUmak9Z1SJosqVPSmKp21kp6Qy+HMbMSyD0Q5pFTLGkXsrmMZwKHAFMA0rSda9L7WWTzFs9OKXrjI2INWVbJ0amdGUBnRDzV0C/FzFoq90BIPjnFM4BFEfF0esh6YVX7HWQPXM8BvpyOPQ1YntYvJAVdsiyX6n27OdfYrDyKEAjzyinu7ZiVnOLpZKPKMWQz1VUySu4CJkkaB3wAuK5mh51rbFYaRQiEeeQULwPmShoraRhwfFU7y4ADga4UYFcDp5KqzESWinM98FXgwUqgNbPyyv2ucV45xZIWkI3ungRWAUNTf16S9BiwNLXRAZwI3Fd1rIVkp8rz6vmMzjU2KzbnGreAc43NGsu5xmZmDeZAaGZtz4HQzNqeA6GZtb1BHQibUXRB0gJJn2lsT80sT7k/PtMCLrpgZn0qxIiwREUXJqXdp0halLY9oxnfiZm1Tu6BsGRFF9amNncjS+ebDpybslPMrKRyD4SUr+gCwE0pj3kDsL6qH91cdMGsPIoQCMtWdKFnHzZT41qriy6YlUcRAmGpii6Y2eCT+13jkhZdGBAXXTArNhddaAEXXTBrLBddMDNrMAdCM2t7DoRm1vYcCKk/Jzlt2ylpp1b2z8yaK/e7xgXSb05y5VGcgeot17jRnLtstnVKNSLMKSd5o6QvSlpGlvECcLqkVZLuk7Rbsz6vmbVGaQJhHjnJySjg/oiYERGVYd2GiNgXuBBwSS6zkitNICSfnGTIHsS+tkdfKnMZrwQm1uqsc43NyqNMgTCvnORNqSpNtUrbNfOMUzvONTYriTIFwjxyks2sDZTmrnFeOcmN4Fxjs2JzrnELONfYrLGca2xm1mAOhGbW9hwIzaztORCaWdtzIDSztlfKQChpgaRcUtskzZe0VlLUW4XmkQ0bm90tM9sGpQyEObsDOBh4NO+OmFljlCYQSjpL0kOS/gv4q7TsLyX9QtJKSR2VSjB9VJ/ZRdKSVHPw/qoJ3w+VdFeqKHO1pNG99SMi7omIzuZ/YjNrlVIEQkn7kaXS7QMcQzbhOsBFwOkRsR9ZFZhvV+1Wq/rMScDNqe7gXsDqdHr7f4CDU0WZFWSFF7a1zy66YFYSZUmxmw1cHxF/BJB0IzCCbP7hqyVVtntN1T4/jogu4NeSKtVnlpOl5A1L61dLeicwBbgjtTOcbJrPbRIRF5EFaobt8kan75gVWFkCIby68swQ4L/7qCpdq/rMEklzgKOASyX9X+BZ4JaIOLHRHTazcijFqTGwBDha0vaSdgDeC/wReETS8QDK7NVXI5ImAOsj4nvAxcC+ZBO5z5Q0KW0zUtLkPpoZsF136vWSo5kVQCkCYUSsAhYCq8mKpHakVR8CTpH0K+AB4P39NDWX7LrgPcCxwNcj4mlgHnCFpHvJAmOv5fclnSHpcWA8cK+k72/t5zKzYnD1mRZw9RmzxnL1GTOzBivTzZKWknQ9sGuPxZ+NiJvz6I+ZNY8DYS8i4ui8+2BmrVGIU2NJcyUdWPX+NEkf3sq25kl6Y9X770ua0oh+pvbmpAyUlysZK/1xrrFZsRVlRDgX2Eg2GRMR8Z1taGsecD/wRGrr49vYt55+l47h+YzNBommjggl/TjlAT8g6RNp2eFpRPUrSbdKmgicBvx9ygGeXakuI+ntku6uam9iesQFSedIWp5yhi9KzxEeB0wFLk9tbS9pkaSpaZ8TJd2X9jm/qt2Nks5LfVpalYnyKhHRGRH3Al1N+MrMLAfNPjX+WMoDngqckQLM94BjI2Iv4PhUwOA7wNciYu+IqDwjSEQ8CAxPU2xCNhfxVen1BRExLSL2ALYH3hMR15DlCn8otfVipa10unw+2UTxewPTJH0grR4FLE19WgL8r2394M41NiuPZgfCM9LDzkuBNwOfAJZExCMAdU6ZeRXw1+n1CWQPVgO8S9IySfeRBbfd+2lnGrAoIp6OiJeBy8mm8QT4E/DT9HolMLGOfvXJE7yblUfTAqGkuWR1+w5II617gF/x6pzh/iwkm9h9MhAR8bCkEWSVZo6LiD3JRpkj+utSH+v+HFueLN9Mca6dmlkLNHNE+Drg2Yj4Y6oTuD9ZdZh3StoVQNKOadvngR1qNRIR68iC09lsGQ1Wgt6GVDuw+u5tb20tS8feSdJQ4ERg8dZ+ODMbPJoZCH8BbJdubvwT2enx02Snx9elU+ZKYPsJWVGF1ZViqT0sBE4mXR+MiP8mGwXeB/yYrLxWxY+A71RullQWRsSTwJnAL8lGpqsi4oaBfihJ01Ku8fHAdyU90N8+LrpgVmzONW4B5xqbNZZzjc3MGsw3BXoh6Syy099qV0fEeXn0x8yax4GwFyngNS/obYYhG4egzUJ93tCjrgGGAAAM9ElEQVS2IguCGBp0je6CoXn3xraWA2FOhmwcws6v35nXvf51VM25YiUTETz37HOsf3Y9Xa9zslFZlfIaofKd4P1yZdOK3i+pMhHUwNvZLAfBQUBKf8fN/juWWSkDYc4uJyvlvydZat9WFXUQchAcJCRf3ii70gRCFWeC959FAtxNNneJmZVYKa4R6pUTvG8HrCLLCb4IOC2l3c0gS7s7KO1WmeB9N+BG4Bq2TPB+XsouGalXTvD+gqTPkk3w/sV++jQM+Bvg73pZ/wmyh8fZ+Y079/sZdz9nUb/bDMQDX5zb0PaWLF7C8OHD2f+A/QH4/kXfZ+TIkZx08kkDbuuyf7+Mdx/8bnZ54y4AfOq0TzH/7+bz9re/vaF9NqtXKQIhxZzg/dtkBSQ6aq2snuB98p6TS//UeseSDkaPGt0dCD/+ia0v83jZpZcxZfcp3YHwW9/5VkP6WK+XX36Z7bYryz99a4XSnBrTxwTvVT/VQ4qaE7yTVZz5PdkE7x9O626pamNKRJzSV0cknQuMIxs5ltoHj/sgs/afxdS9p/KD7/8AgFtuvoWZM2ay/9T9Oeqwo3i081Eu/t7FXPDNCzhg2gHccfsdnPdP5/H1r36d3zz4G945853d7T3a+Sgz9psBwJfP+zJzDpzDtH2mMf+T84kIrr/ueu5ZeQ+nfOQUDph2AC+++CKHH3I4q1auAuCqhVcxfd/pTNtnGmd//uzudt+w4xtYcM4C9p+6P++a/S6eeuopAK679jqm7TON/afuz6HvPrTXz3nZv1/GySeezPFHH8/7jnofGzdu5KjDjmLmjJlM33c6P73xp93932fPffjUaZ9i2j7T+NhHPsYvb/0lB889mL2m7MWK5Ssa+wewQihLICzMBO+SPg4cBpyYRpyl9u2Lvs3tS2+n464OLvzWhTz11FPM/9v5XH7l5SxdsZRLr7iUCRMncMr/OoX5p8/nruV3MXPWzO79d3v7bvzpT3/ikd8+AsC111zLMccdA8CpnzyVJXcuYfk9y9m0aRM/v+nnHH3M0eyz3z5cfMnF3LX8LrbfvjsdnCefeJJzzjqHm26+ibuW38XKlSv5yQ0/AeCFF15g+vTpLF2xlJmzZvKjH/wIgH8+75+54ac3sHTFUq669ir6cveyu/nuxd/lZzf/jBEjRnDF1Vdwx7I7+Nl//ozPf/bzVNJNf7vut3xy/idZtnIZax5aw1ULr+KWX97Cef98Hv96/r827Lu34ihFICzSBO9kRWTfANyVbrqcs1UfqiAu/NaF3aOs3z/+e3548Q+ZOWsmE3edCMCOO+7Y5/4Axxx3DNddex0A1159LccedyyQXVecO2su0/edzuJFi3nwwQf7bGflipXMnjObcePGsd1223HCB0/gjtvvAGD48OEccdQRAOy979787tHfAbD/Aftz6sdP5YcX/5DNmzf32f5B7z6o+/NEBAvOXsCM/Wbw3iPeyxNPPMH6p9YDMHHiRPbYYw+GDBnC26e8nbnvmoskdt9jdx599NF+vw8rn9JcKOkj0+NVmdcRMa/H+9Hp9yXAJTW2v42scGs9/SjNd9afJYuXsOi2Rdy25DZGjhzJ4Ycczp577snDax4eUDvHHncsf3PS3/C+978PSUx62yQ2bdrE35/x93Tc2cH4N4/nvH86j02bNvXZTl8FQIYNG9b9uNHQoUN5+eWXAfjGt77B8ruX84uf/4IDpx/InXffydixY2u2MbKqQO7CKxayYcMGbl96O8OGDWPK5Cnd/Rv+muHd2w0ZMqT7/ZAhQ7qPa4NLKUaE1hz/89z/MGbMGEaOHMlDv3mI5cuW89KfXuL2jtvpfKQTgGeeyYqI7zB6B57f+HzNdt76l29l6NChnP/l87tHg5WgMnansWzcuJEfX/fj7u13GL0Dzz//6ramTZ/G7R23s2HDBjZv3szVC69m1uxZfX6G3677LdOmT+Psc89m7Nix/P7x39f12Z977jnGjRvHsGHDWLxocfcI09rToBndNJpaPMF7ox93qcchhx3Cxd+7mBn7zeBtk9/GtBnT2Gmnnfjmt77JSSecRFdXF+PGjeMnP/8JRxx1BCefeDI3/eQm/vVrr75Oduxxx3LWmWfxwENZecYxY8Yw72PzmLHvDN4y4S3sN3W/7m0/9OEP8en5n2bE9iO4bclt3cv/Ype/4Av/9AWOPPRIIoLDDj+M97zvPX1+hrPOPIt1a9cREcx911z2fMeedX32E048gb8+5q+ZfcBs3rHXO5j8V71eFrY24HqELVCrHuF2z2zHpMmTcuqRNdraNWt5eUefNrfKoKxHqHJN8P4Pkn4t6V5l05FOaFTbZpaPopwaz6U8E7zfA0xNc7F8EvgXstn1rAD+6z//i7PPOvsVyyZMnMCVV1+ZU4+sDJoaCCX9mGwazxFkj6pcJOlw4Etk1ds2AKeQTfC+WdLJwOnAu8kC403AJRExPbU3EbgxIt6RHlt5L1nhgzuBU8keialM8P4icADwc+AzEbFC0onA58keor4pIj6b2t0IfB14D/Ai8P6IeKrWZ4qIX1a9XUo2l4oVxMGHHszBhx6cdzesZDzBe2ZrJ3g/hSzQDlgQfT4uYuUREcSAZ6m1IvEE75kBT/CeRq9Tgf/by/pPSFohacVzzzz3qvUxNCvo6WBYbpXCrDHUf8cya9qpcY8J3v8oaRHZNJp/NcCmFpIVVriOV0/wPjUiHpO0gBZO8C7pYOAs4J0R8VKtbforutA1uov1z67n6Q1Pu5Zdib2iVL+VVjOvEfY5wXtEPCJpxzQqfB54ba1GImKdpHomeL8mLetrgvevp7Jbz5JN8P7NgX4oSfsA3wUOj4j1A92/21Bc2t2sIJoZCH8BnJbydx/i1RO8DwHWA4eQTfB+jaT3k90s6Wkh2SnorpBN8C6pMsF7J7UneK/cLCHt86SkygTvAn62NRO8p36MZkv5r99FxPu2oh0zKwg/UN0CnuDdrLEG5QPVZmZ58oiwF2rgBO+Snie7PJCnncie28xbEfrhPrRPHyZExLj+NnIgbAFJK+oZng/2PhSlH+6D+9CTT43NrO05EJpZ23MgbI2L8u4AxegDFKMf7kPGfUh8jdDM2p5HhGbW9hwIm0zS4ZIekrRW0ue2sa0fSFov6f6qZTtKukXSw+n366vWnZmO+5Ckw6qW7yfpvrTuG0opMpJeI2lhWr4slT3r2Yc3S/qlpAclPSDp73LqxwhJd0v6VerHF3Lqx1BJ90j6aR7HT9t1pv1XS1qR0/cwRtI1kn6T/m0ckMd3sdUiwj9N+iGrubgOeCswnKzoxJRtaG8O2VzM91ct+xfgc+n154Dz0+sp6XivIUtNXAcMTevuJks/FFkZsSPS8r8FvpNefxBYWKMPuwD7ptc7AGvSsVrdDwGj0+thZLnk++fQj38A/gP4aR5/j7SuE9ipx7JWfw+XAB9Pr4cDY/L4Lrb6v628g8Vg/kl/0Jur3p8JnLmNbU7klYHwIWCX9HoX4KFaxwJuTv3ZBfhN1fITge9Wb5Neb0f2oKv66c8NZPniufUDGAmsAma0sh/AeOBWsjJwlUDY8u+B2oGwld/Da4FHevYtz38TA/3xqXFzvQl4rOr942lZI70hIp6ErLAEsHM/x35Tel2rT937RFaz8Tmg9iTBdFcM34dsNNbyfqTT0tVkxTtuiYhW9+PfgP8NVJcRyuPvEcB/Slop6RM59OOtZAVVfpguE3xf0qicvout4kDYXLUKDbbqNn1vx+6rT3X3V1n5s2uBT0fE/+TRj4jYHBF7k43Mpkvao1X9kPQeYH1ErOzjmE07fg8zI2Jf4AjgU5Lm1Nimmf3YjuySzYURsQ/wAtmpcCv7sE0cCJvrcbLK3BXjSZNKNdBTknYBSL8rNRJ7O/bj6XWtPnXvI2k7spqSr6oiLmkYWRC8PCKuy6sfFRHx38Ai4PAW9mMm8D5JncCVwEGSLsvje4iIykRl64Hrgekt7sfjwONpRA5ZbdB98/gutpYDYXMtB94maVdJw8ku8t7Y4GPcCHwkvf4I2TW7yvIPprttuwJvA+5OpyjPS9o/3ZH7cI99Km0dB9wW6aJMRdrnYuDBiPhqjv0YJ2lMer09WTX037SqHxFxZkSMj4iJZH/X2yLi5By+h1GSdqi8Bg4lm8WxZf2IiP8HPCapUn3+3cCvW/1dbJNGXWz0T68X8o8ku7O6DjhrG9u6AngS+DPZ/yFPIbtOcivwcPq9Y9X2Z6XjPkS6+5aWTyX7j2UdcAFbHqwfAVwNrCW7e/fWGn2YRXZKci+wOv0cmUM/3kE2teq9qY1z0vKW9iNtN5ctN0ta/T28lewO7K+AByr/xnLox95kE6fdC/wYeH0ef4ut/XFmiZm1PZ8am1nbcyA0s7bnQGhmbc+B0MzangOhmbU9B0Iza3sOhGbW9hwIzazt/f+h+zNQ1CRV3wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 288x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_layers_ram(layers_ram, ax=None, max_ram=64e3):\n",
    "    \n",
    "    if not ax:\n",
    "        fig, ax = plt.subplots(1, figsize=(4, 6))\n",
    "    \n",
    "    l = layers_ram.sort_index(ascending=False)\n",
    "    l['activations_ram'] = 4 * l.activations\n",
    "    l = l[l.activations_ram.notna()]\n",
    "    l.plot(kind='barh', ax=ax, y='activations_ram', x='name')\n",
    "    ax.axvspan(xmin=0, xmax=max_ram, alpha=0.2, color='green')\n",
    "    \n",
    "    return fig\n",
    "    \n",
    "plot_layers_ram(ll);\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(None, 60, 31, 1)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.layers[0].get_output_shape_at(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: calculate per-class accuracy. For each fold\n",
    "# TODO: compare single-frame accuracy (from training) versus voted accuracy\n",
    "# TODO: compare no-overlap, 50% overlap and 10% overlap"
   ]
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}