us8k_cnn_salamon.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "i7x6wKsIGKLK"
   },
   "source": [
    "**Reproduction of:**\n",
    "\n",
    "**Deep Convolutional Neural Networks and Data Augmentation for Environmental  Sound Classification** \n",
    "\n",
    "Justin Salamon and Juan Pablo Bello\n",
    "\n",
    "Some code taken from https://github.com/jaron/deep-listening/blob/master/4-us8k-cnn-salamon.ipynb"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "jNGuDtV8G3v9"
   },
   "source": [
    "# Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#import os\n",
    "\n",
    "#for dir in [load_dir, augmented_load_dir]:\n",
    "#    for k in range(1,10+1):\n",
    "#        filename = \"fold\"+str(k)+\"_x.npy\"\n",
    "#        file_path = os.path.join(dir,filename)\n",
    "#        file= np.load(file_path, allow_pickle = True)\n",
    "#        file = file.astype('float32') \n",
    "#        np.save(file_path, file, allow_pickle = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!git clone https://github.com/grudloff/Salomon2017Replication"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Kj7xR5hfzYAW"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import math\n",
    "\n",
    "from preprocessing_augmented import load_folds\n",
    "from model import build_model\n",
    "from evaluation import evaluate\n",
    "\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras.utils import Sequence\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "import seaborn as sn\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "load_dir = \"CNN-Sound/data/us8k\"\n",
    "augmented_load_dir = \"CNN-Sound/data/us8k-augmented\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "QLYCNfTzwWBt"
   },
   "source": [
    "10-Fold Crossvalidation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 555
    },
    "colab_type": "code",
    "id": "WQyEXvd9wWBt",
    "outputId": "78f34333-1791-4de4-9206-1d2cf9591893"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "*** Train on {3, 4, 5, 6, 7, 8, 9} Validate on 1 Test on 2 ***\n",
      "val shape:  (873, 128, 128, 1)\n",
      "test shape:  (888, 128, 128, 1)\n",
      "train shape:  (146391, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 146391 samples, validate on 888 samples\n",
      "Epoch 1/100\n",
      "146391/146391 [==============================] - 89s 607us/sample - loss: 2.5204 - accuracy: 0.2183 - val_loss: 2.0363 - val_accuracy: 0.3187\n",
      "Epoch 2/100\n",
      "146391/146391 [==============================] - 85s 578us/sample - loss: 1.9122 - accuracy: 0.3899 - val_loss: 1.6084 - val_accuracy: 0.4606\n",
      "Epoch 3/100\n",
      "146391/146391 [==============================] - 85s 579us/sample - loss: 1.6812 - accuracy: 0.4762 - val_loss: 1.4879 - val_accuracy: 0.5113\n",
      "Epoch 4/100\n",
      "146391/146391 [==============================] - 85s 579us/sample - loss: 1.5428 - accuracy: 0.5238 - val_loss: 1.4137 - val_accuracy: 0.5495\n",
      "Epoch 5/100\n",
      "146391/146391 [==============================] - 86s 587us/sample - loss: 1.4491 - accuracy: 0.5564 - val_loss: 1.3849 - val_accuracy: 0.5507\n",
      "Epoch 6/100\n",
      "146391/146391 [==============================] - 85s 583us/sample - loss: 1.3645 - accuracy: 0.5840 - val_loss: 1.2603 - val_accuracy: 0.5833\n",
      "Epoch 7/100\n",
      "146391/146391 [==============================] - 85s 579us/sample - loss: 1.2941 - accuracy: 0.6055 - val_loss: 1.2667 - val_accuracy: 0.5845\n",
      "Epoch 8/100\n",
      "146391/146391 [==============================] - 85s 578us/sample - loss: 1.2360 - accuracy: 0.6229 - val_loss: 1.3697 - val_accuracy: 0.5518\n",
      "Epoch 9/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 1.1917 - accuracy: 0.6355 - val_loss: 1.1772 - val_accuracy: 0.6070\n",
      "Epoch 10/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 1.1471 - accuracy: 0.6497 - val_loss: 1.1959 - val_accuracy: 0.6160\n",
      "Epoch 11/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 1.1089 - accuracy: 0.6603 - val_loss: 1.2082 - val_accuracy: 0.5957\n",
      "Epoch 12/100\n",
      "146391/146391 [==============================] - 85s 580us/sample - loss: 1.0783 - accuracy: 0.6686 - val_loss: 1.1776 - val_accuracy: 0.6261\n",
      "Epoch 13/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 1.0516 - accuracy: 0.6770 - val_loss: 1.1861 - val_accuracy: 0.6025\n",
      "Epoch 14/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 1.0188 - accuracy: 0.6861 - val_loss: 1.1593 - val_accuracy: 0.6340\n",
      "Epoch 15/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.9936 - accuracy: 0.6940 - val_loss: 1.2080 - val_accuracy: 0.6216\n",
      "Epoch 16/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.9687 - accuracy: 0.6996 - val_loss: 1.1971 - val_accuracy: 0.6441\n",
      "Epoch 17/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.9496 - accuracy: 0.7049 - val_loss: 1.1525 - val_accuracy: 0.6273\n",
      "Epoch 18/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.9352 - accuracy: 0.7098 - val_loss: 1.1068 - val_accuracy: 0.6374\n",
      "Epoch 19/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.9074 - accuracy: 0.7173 - val_loss: 1.2135 - val_accuracy: 0.6025\n",
      "Epoch 20/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.8902 - accuracy: 0.7229 - val_loss: 1.1416 - val_accuracy: 0.6340\n",
      "Epoch 21/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.8799 - accuracy: 0.7259 - val_loss: 1.2186 - val_accuracy: 0.6036\n",
      "Epoch 22/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.8662 - accuracy: 0.7283 - val_loss: 1.2175 - val_accuracy: 0.6475\n",
      "Epoch 23/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.8523 - accuracy: 0.7337 - val_loss: 1.1964 - val_accuracy: 0.6396\n",
      "Epoch 24/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.8360 - accuracy: 0.7371 - val_loss: 1.1606 - val_accuracy: 0.6554\n",
      "Epoch 25/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.8276 - accuracy: 0.7396 - val_loss: 1.1898 - val_accuracy: 0.6396\n",
      "Epoch 26/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.8174 - accuracy: 0.7417 - val_loss: 1.2491 - val_accuracy: 0.6363\n",
      "Epoch 27/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.8097 - accuracy: 0.7457 - val_loss: 1.2424 - val_accuracy: 0.6363\n",
      "Epoch 28/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.7955 - accuracy: 0.7483 - val_loss: 1.2321 - val_accuracy: 0.6475\n",
      "Epoch 29/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.7970 - accuracy: 0.7470 - val_loss: 1.1781 - val_accuracy: 0.6577\n",
      "Epoch 30/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.7810 - accuracy: 0.7511 - val_loss: 1.2955 - val_accuracy: 0.5991\n",
      "Epoch 31/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.7743 - accuracy: 0.7545 - val_loss: 1.0986 - val_accuracy: 0.6757\n",
      "Epoch 32/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.7662 - accuracy: 0.7568 - val_loss: 1.1334 - val_accuracy: 0.6577\n",
      "Epoch 33/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.7553 - accuracy: 0.7598 - val_loss: 1.1884 - val_accuracy: 0.6599\n",
      "Epoch 34/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.7568 - accuracy: 0.7595 - val_loss: 1.2809 - val_accuracy: 0.6464\n",
      "Epoch 35/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.7483 - accuracy: 0.7612 - val_loss: 1.3567 - val_accuracy: 0.6115\n",
      "Epoch 36/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.7382 - accuracy: 0.7648 - val_loss: 1.2598 - val_accuracy: 0.6351\n",
      "Epoch 37/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.7348 - accuracy: 0.7653 - val_loss: 1.2304 - val_accuracy: 0.6430\n",
      "Epoch 38/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.7297 - accuracy: 0.7677 - val_loss: 1.1935 - val_accuracy: 0.6588\n",
      "Epoch 39/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.7243 - accuracy: 0.7691 - val_loss: 1.2355 - val_accuracy: 0.6194\n",
      "Epoch 40/100\n",
      "146391/146391 [==============================] - 84s 576us/sample - loss: 0.7181 - accuracy: 0.7705 - val_loss: 1.3140 - val_accuracy: 0.6441\n",
      "Epoch 41/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.7115 - accuracy: 0.7737 - val_loss: 1.2963 - val_accuracy: 0.6351\n",
      "Epoch 42/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.7091 - accuracy: 0.7731 - val_loss: 1.2349 - val_accuracy: 0.6520\n",
      "Epoch 43/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.6965 - accuracy: 0.7775 - val_loss: 1.1982 - val_accuracy: 0.6554\n",
      "Epoch 44/100\n",
      "146391/146391 [==============================] - 84s 575us/sample - loss: 0.6955 - accuracy: 0.7773 - val_loss: 1.3168 - val_accuracy: 0.6430\n",
      "Epoch 45/100\n",
      "146391/146391 [==============================] - 85s 577us/sample - loss: 0.6846 - accuracy: 0.7799 - val_loss: 1.3587 - val_accuracy: 0.6306\n",
      "Epoch 46/100\n",
      "146391/146391 [==============================] - 84s 577us/sample - loss: 0.6826 - accuracy: 0.7814 - val_loss: 1.2195 - val_accuracy: 0.6610\n",
      "Evaluating model...\n",
      "ROC: 0.941\n",
      "\n",
      "Accuracy = 0.65\n",
      "F-Score: 0.65\n",
      "\n",
      "*** Train on {1, 3, 4, 5, 7, 8, 9} Validate on 2 Test on 6 ***\n",
      "val shape:  (888, 128, 128, 1)\n",
      "test shape:  (823, 128, 128, 1)\n",
      "train shape:  (147441, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 147441 samples, validate on 823 samples\n",
      "Epoch 1/100\n",
      "147441/147441 [==============================] - 87s 589us/sample - loss: 2.0670 - accuracy: 0.3021 - val_loss: 1.9366 - val_accuracy: 0.3244\n",
      "Epoch 2/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 1.5762 - accuracy: 0.4830 - val_loss: 1.9714 - val_accuracy: 0.3463\n",
      "Epoch 3/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 1.3704 - accuracy: 0.5641 - val_loss: 1.6007 - val_accuracy: 0.4520\n",
      "Epoch 4/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "147441/147441 [==============================] - 85s 574us/sample - loss: 1.2308 - accuracy: 0.6132 - val_loss: 1.5986 - val_accuracy: 0.4386\n",
      "Epoch 5/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 1.1292 - accuracy: 0.6489 - val_loss: 1.3295 - val_accuracy: 0.5541\n",
      "Epoch 6/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 1.0421 - accuracy: 0.6793 - val_loss: 1.3597 - val_accuracy: 0.5589\n",
      "Epoch 7/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.9795 - accuracy: 0.6990 - val_loss: 1.2444 - val_accuracy: 0.5966\n",
      "Epoch 8/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 0.9258 - accuracy: 0.7148 - val_loss: 1.3472 - val_accuracy: 0.5978\n",
      "Epoch 9/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 0.8860 - accuracy: 0.7273 - val_loss: 1.4694 - val_accuracy: 0.5723\n",
      "Epoch 10/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.8551 - accuracy: 0.7370 - val_loss: 1.4020 - val_accuracy: 0.5735\n",
      "Epoch 11/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 0.8219 - accuracy: 0.7461 - val_loss: 1.4258 - val_accuracy: 0.5662\n",
      "Epoch 12/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.7966 - accuracy: 0.7542 - val_loss: 1.2154 - val_accuracy: 0.6063\n",
      "Epoch 13/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.7711 - accuracy: 0.7623 - val_loss: 1.2778 - val_accuracy: 0.6112\n",
      "Epoch 14/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.7488 - accuracy: 0.7689 - val_loss: 1.2860 - val_accuracy: 0.6148\n",
      "Epoch 15/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.7285 - accuracy: 0.7733 - val_loss: 1.3164 - val_accuracy: 0.5990\n",
      "Epoch 16/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.7082 - accuracy: 0.7802 - val_loss: 1.3250 - val_accuracy: 0.6051\n",
      "Epoch 17/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.6939 - accuracy: 0.7849 - val_loss: 1.2562 - val_accuracy: 0.6027\n",
      "Epoch 18/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.6761 - accuracy: 0.7879 - val_loss: 1.1826 - val_accuracy: 0.6537\n",
      "Epoch 19/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.6597 - accuracy: 0.7948 - val_loss: 1.2033 - val_accuracy: 0.6428\n",
      "Epoch 20/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 0.6461 - accuracy: 0.7988 - val_loss: 1.3758 - val_accuracy: 0.6294\n",
      "Epoch 21/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 0.6344 - accuracy: 0.8020 - val_loss: 1.6149 - val_accuracy: 0.5893\n",
      "Epoch 22/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.6205 - accuracy: 0.8063 - val_loss: 1.2477 - val_accuracy: 0.6391\n",
      "Epoch 23/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.6094 - accuracy: 0.8097 - val_loss: 1.2566 - val_accuracy: 0.6476\n",
      "Epoch 24/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5992 - accuracy: 0.8125 - val_loss: 1.2127 - val_accuracy: 0.6561\n",
      "Epoch 25/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5876 - accuracy: 0.8163 - val_loss: 1.3096 - val_accuracy: 0.6428\n",
      "Epoch 26/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5743 - accuracy: 0.8213 - val_loss: 1.3057 - val_accuracy: 0.6416\n",
      "Epoch 27/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5696 - accuracy: 0.8224 - val_loss: 1.3653 - val_accuracy: 0.6525\n",
      "Epoch 28/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5613 - accuracy: 0.8243 - val_loss: 1.3831 - val_accuracy: 0.6731\n",
      "Epoch 29/100\n",
      "147441/147441 [==============================] - 85s 576us/sample - loss: 0.5489 - accuracy: 0.8289 - val_loss: 1.3824 - val_accuracy: 0.6561\n",
      "Epoch 30/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5413 - accuracy: 0.8313 - val_loss: 1.3256 - val_accuracy: 0.6501\n",
      "Epoch 31/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5320 - accuracy: 0.8335 - val_loss: 1.3506 - val_accuracy: 0.6768\n",
      "Epoch 32/100\n",
      "147441/147441 [==============================] - 85s 574us/sample - loss: 0.5269 - accuracy: 0.8352 - val_loss: 1.2535 - val_accuracy: 0.6756\n",
      "Epoch 33/100\n",
      "147441/147441 [==============================] - 85s 575us/sample - loss: 0.5212 - accuracy: 0.8370 - val_loss: 1.2809 - val_accuracy: 0.6634\n",
      "Evaluating model...\n",
      "ROC: 0.946\n",
      "\n",
      "Accuracy = 0.66\n",
      "F-Score: 0.66\n",
      "\n",
      "*** Train on {1, 2, 4, 6, 7, 8, 9} Validate on 3 Test on 5 ***\n",
      "val shape:  (925, 128, 128, 1)\n",
      "test shape:  (936, 128, 128, 1)\n",
      "train shape:  (144291, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 144291 samples, validate on 936 samples\n",
      "Epoch 1/100\n",
      "144291/144291 [==============================] - 81s 564us/sample - loss: 2.2024 - accuracy: 0.2407 - val_loss: 1.5955 - val_accuracy: 0.4765\n",
      "Epoch 2/100\n",
      "144291/144291 [==============================] - 81s 563us/sample - loss: 1.5326 - accuracy: 0.5083 - val_loss: 1.4064 - val_accuracy: 0.5662\n",
      "Epoch 3/100\n",
      "144291/144291 [==============================] - 82s 567us/sample - loss: 1.2560 - accuracy: 0.6148 - val_loss: 1.1359 - val_accuracy: 0.6635\n",
      "Epoch 4/100\n",
      "144291/144291 [==============================] - 82s 569us/sample - loss: 1.1092 - accuracy: 0.6664 - val_loss: 1.1312 - val_accuracy: 0.6677\n",
      "Epoch 5/100\n",
      "144291/144291 [==============================] - 81s 563us/sample - loss: 1.0144 - accuracy: 0.6951 - val_loss: 1.0767 - val_accuracy: 0.6741\n",
      "Epoch 6/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.9535 - accuracy: 0.7136 - val_loss: 1.0527 - val_accuracy: 0.6923\n",
      "Epoch 7/100\n",
      "144291/144291 [==============================] - 81s 562us/sample - loss: 0.8925 - accuracy: 0.7334 - val_loss: 0.9695 - val_accuracy: 0.7094\n",
      "Epoch 8/100\n",
      "144291/144291 [==============================] - 81s 562us/sample - loss: 0.8491 - accuracy: 0.7451 - val_loss: 1.0444 - val_accuracy: 0.7105\n",
      "Epoch 9/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.8115 - accuracy: 0.7551 - val_loss: 0.9809 - val_accuracy: 0.7233\n",
      "Epoch 10/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.7772 - accuracy: 0.7660 - val_loss: 1.0298 - val_accuracy: 0.7190\n",
      "Epoch 11/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.7495 - accuracy: 0.7739 - val_loss: 0.9669 - val_accuracy: 0.7222\n",
      "Epoch 12/100\n",
      "144291/144291 [==============================] - 81s 562us/sample - loss: 0.7218 - accuracy: 0.7821 - val_loss: 0.9702 - val_accuracy: 0.7372\n",
      "Epoch 13/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.7006 - accuracy: 0.7884 - val_loss: 0.9286 - val_accuracy: 0.7350\n",
      "Epoch 14/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.6823 - accuracy: 0.7924 - val_loss: 1.0308 - val_accuracy: 0.7190\n",
      "Epoch 15/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.6610 - accuracy: 0.7983 - val_loss: 0.9737 - val_accuracy: 0.7436\n",
      "Epoch 16/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.6483 - accuracy: 0.8021 - val_loss: 0.9225 - val_accuracy: 0.7511\n",
      "Epoch 17/100\n",
      "144291/144291 [==============================] - 81s 562us/sample - loss: 0.6255 - accuracy: 0.8087 - val_loss: 0.9347 - val_accuracy: 0.7618\n",
      "Epoch 18/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.6117 - accuracy: 0.8121 - val_loss: 0.9566 - val_accuracy: 0.7553\n",
      "Epoch 19/100\n",
      "144291/144291 [==============================] - 81s 562us/sample - loss: 0.5996 - accuracy: 0.8153 - val_loss: 0.8461 - val_accuracy: 0.7468\n",
      "Epoch 20/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.5927 - accuracy: 0.8177 - val_loss: 0.8996 - val_accuracy: 0.7521\n",
      "Epoch 21/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.5749 - accuracy: 0.8224 - val_loss: 0.9913 - val_accuracy: 0.7521\n",
      "Epoch 22/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.5670 - accuracy: 0.8254 - val_loss: 0.9243 - val_accuracy: 0.7639\n",
      "Epoch 23/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.5564 - accuracy: 0.8268 - val_loss: 0.8934 - val_accuracy: 0.7532\n",
      "Epoch 24/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.5483 - accuracy: 0.8298 - val_loss: 1.0047 - val_accuracy: 0.7575\n",
      "Epoch 25/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.5397 - accuracy: 0.8325 - val_loss: 0.9011 - val_accuracy: 0.7425\n",
      "Epoch 26/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.5277 - accuracy: 0.8357 - val_loss: 0.9754 - val_accuracy: 0.7468\n",
      "Epoch 27/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.5220 - accuracy: 0.8372 - val_loss: 0.9105 - val_accuracy: 0.7532\n",
      "Epoch 28/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.5119 - accuracy: 0.8403 - val_loss: 0.9425 - val_accuracy: 0.7543\n",
      "Epoch 29/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.5067 - accuracy: 0.8419 - val_loss: 0.9372 - val_accuracy: 0.7585\n",
      "Epoch 30/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.4979 - accuracy: 0.8442 - val_loss: 0.9620 - val_accuracy: 0.7575\n",
      "Epoch 31/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.4941 - accuracy: 0.8453 - val_loss: 1.0058 - val_accuracy: 0.7607\n",
      "Epoch 32/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.4864 - accuracy: 0.8475 - val_loss: 0.8937 - val_accuracy: 0.7532\n",
      "Epoch 33/100\n",
      "144291/144291 [==============================] - 81s 561us/sample - loss: 0.4800 - accuracy: 0.8498 - val_loss: 0.8736 - val_accuracy: 0.7532\n",
      "Epoch 34/100\n",
      "144291/144291 [==============================] - 81s 560us/sample - loss: 0.4763 - accuracy: 0.8517 - val_loss: 0.9608 - val_accuracy: 0.7489\n",
      "Evaluating model...\n",
      "ROC: 0.928\n",
      "\n",
      "Accuracy = 0.63\n",
      "F-Score: 0.63\n",
      "\n",
      "*** Train on {1, 3, 5, 6, 7, 8, 9} Validate on 4 Test on 2 ***\n",
      "val shape:  (990, 128, 128, 1)\n",
      "test shape:  (888, 128, 128, 1)\n",
      "train shape:  (143934, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 143934 samples, validate on 888 samples\n",
      "Epoch 1/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 2.2050 - accuracy: 0.2661 - val_loss: 1.8076 - val_accuracy: 0.4336\n",
      "Epoch 2/100\n",
      "143934/143934 [==============================] - 81s 562us/sample - loss: 1.5596 - accuracy: 0.4994 - val_loss: 1.5716 - val_accuracy: 0.5101\n",
      "Epoch 3/100\n",
      "143934/143934 [==============================] - 81s 562us/sample - loss: 1.3484 - accuracy: 0.5761 - val_loss: 1.5234 - val_accuracy: 0.5146\n",
      "Epoch 4/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 1.2257 - accuracy: 0.6160 - val_loss: 1.4321 - val_accuracy: 0.5428\n",
      "Epoch 5/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 1.1385 - accuracy: 0.6473 - val_loss: 1.4429 - val_accuracy: 0.5867\n",
      "Epoch 6/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 1.0672 - accuracy: 0.6699 - val_loss: 1.4724 - val_accuracy: 0.5372\n",
      "Epoch 7/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 1.0159 - accuracy: 0.6866 - val_loss: 1.8957 - val_accuracy: 0.5259\n",
      "Epoch 8/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.9682 - accuracy: 0.7000 - val_loss: 1.4821 - val_accuracy: 0.5912\n",
      "Epoch 9/100\n",
      "143934/143934 [==============================] - 81s 560us/sample - loss: 0.9315 - accuracy: 0.7114 - val_loss: 1.4240 - val_accuracy: 0.5811\n",
      "Epoch 10/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.9018 - accuracy: 0.7201 - val_loss: 1.8400 - val_accuracy: 0.5574\n",
      "Epoch 11/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.8705 - accuracy: 0.7298 - val_loss: 1.5110 - val_accuracy: 0.6047\n",
      "Epoch 12/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.8472 - accuracy: 0.7361 - val_loss: 1.3999 - val_accuracy: 0.6047\n",
      "Epoch 13/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.8235 - accuracy: 0.7448 - val_loss: 1.4181 - val_accuracy: 0.5856\n",
      "Epoch 14/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.8074 - accuracy: 0.7481 - val_loss: 1.4093 - val_accuracy: 0.5901\n",
      "Epoch 15/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.7891 - accuracy: 0.7532 - val_loss: 1.5338 - val_accuracy: 0.5980\n",
      "Epoch 16/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.7680 - accuracy: 0.7592 - val_loss: 1.4962 - val_accuracy: 0.5721\n",
      "Epoch 17/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.7502 - accuracy: 0.7640 - val_loss: 1.3918 - val_accuracy: 0.6002\n",
      "Epoch 18/100\n",
      "143934/143934 [==============================] - 81s 560us/sample - loss: 0.7383 - accuracy: 0.7685 - val_loss: 1.5404 - val_accuracy: 0.6126\n",
      "Epoch 19/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.7179 - accuracy: 0.7745 - val_loss: 1.5056 - val_accuracy: 0.6014\n",
      "Epoch 20/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.7045 - accuracy: 0.7782 - val_loss: 1.4179 - val_accuracy: 0.5935\n",
      "Epoch 21/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6926 - accuracy: 0.7818 - val_loss: 1.4948 - val_accuracy: 0.6284\n",
      "Epoch 22/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6804 - accuracy: 0.7846 - val_loss: 1.3682 - val_accuracy: 0.6261\n",
      "Epoch 23/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6718 - accuracy: 0.7880 - val_loss: 1.3639 - val_accuracy: 0.6273\n",
      "Epoch 24/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6581 - accuracy: 0.7928 - val_loss: 1.3319 - val_accuracy: 0.6171\n",
      "Epoch 25/100\n",
      "143934/143934 [==============================] - 81s 562us/sample - loss: 0.6480 - accuracy: 0.7948 - val_loss: 1.4752 - val_accuracy: 0.5912\n",
      "Epoch 26/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6416 - accuracy: 0.7960 - val_loss: 1.5117 - val_accuracy: 0.6182\n",
      "Epoch 27/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6271 - accuracy: 0.8002 - val_loss: 1.4869 - val_accuracy: 0.6363\n",
      "Epoch 28/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6211 - accuracy: 0.8022 - val_loss: 1.4307 - val_accuracy: 0.6216\n",
      "Epoch 29/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6130 - accuracy: 0.8055 - val_loss: 1.6253 - val_accuracy: 0.6081\n",
      "Epoch 30/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.6037 - accuracy: 0.8096 - val_loss: 1.4402 - val_accuracy: 0.6498\n",
      "Epoch 31/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5981 - accuracy: 0.8100 - val_loss: 1.3953 - val_accuracy: 0.6464\n",
      "Epoch 32/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5853 - accuracy: 0.8134 - val_loss: 1.5986 - val_accuracy: 0.6284\n",
      "Epoch 33/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5815 - accuracy: 0.8146 - val_loss: 1.4153 - val_accuracy: 0.6351\n",
      "Epoch 34/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5748 - accuracy: 0.8173 - val_loss: 1.3539 - val_accuracy: 0.6351\n",
      "Epoch 35/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5650 - accuracy: 0.8205 - val_loss: 1.3190 - val_accuracy: 0.6419\n",
      "Epoch 36/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5567 - accuracy: 0.8230 - val_loss: 1.3915 - val_accuracy: 0.6396\n",
      "Epoch 37/100\n",
      "143934/143934 [==============================] - 81s 561us/sample - loss: 0.5524 - accuracy: 0.8245 - val_loss: 1.4045 - val_accuracy: 0.6554\n",
      "Epoch 38/100\n",
      "143934/143934 [==============================] - 81s 562us/sample - loss: 0.5485 - accuracy: 0.8254 - val_loss: 2.2407 - val_accuracy: 0.5315\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 39/100\n",
      "143934/143934 [==============================] - 84s 583us/sample - loss: 0.5435 - accuracy: 0.8265 - val_loss: 1.7315 - val_accuracy: 0.6171\n",
      "Epoch 40/100\n",
      "143934/143934 [==============================] - 84s 580us/sample - loss: 0.5333 - accuracy: 0.8292 - val_loss: 1.4263 - val_accuracy: 0.6509\n",
      "Epoch 41/100\n",
      "143934/143934 [==============================] - 84s 581us/sample - loss: 0.5297 - accuracy: 0.8298 - val_loss: 1.6215 - val_accuracy: 0.6408\n",
      "Epoch 42/100\n",
      "143934/143934 [==============================] - 84s 581us/sample - loss: 0.5264 - accuracy: 0.8327 - val_loss: 1.4954 - val_accuracy: 0.6554\n",
      "Epoch 43/100\n",
      "143934/143934 [==============================] - 84s 582us/sample - loss: 0.5164 - accuracy: 0.8355 - val_loss: 1.5169 - val_accuracy: 0.6306\n",
      "Epoch 44/100\n",
      "143934/143934 [==============================] - 84s 580us/sample - loss: 0.5146 - accuracy: 0.8358 - val_loss: 1.6674 - val_accuracy: 0.6329\n",
      "Epoch 45/100\n",
      "143934/143934 [==============================] - 84s 581us/sample - loss: 0.5129 - accuracy: 0.8370 - val_loss: 1.4129 - val_accuracy: 0.6543\n",
      "Epoch 46/100\n",
      "143934/143934 [==============================] - 83s 580us/sample - loss: 0.5069 - accuracy: 0.8375 - val_loss: 1.6099 - val_accuracy: 0.6543\n",
      "Epoch 47/100\n",
      "143934/143934 [==============================] - 83s 580us/sample - loss: 0.5049 - accuracy: 0.8392 - val_loss: 1.4987 - val_accuracy: 0.6520\n",
      "Epoch 48/100\n",
      "143934/143934 [==============================] - 83s 579us/sample - loss: 0.5002 - accuracy: 0.8412 - val_loss: 1.6826 - val_accuracy: 0.6318\n",
      "Epoch 49/100\n",
      "143934/143934 [==============================] - 83s 579us/sample - loss: 0.4942 - accuracy: 0.8418 - val_loss: 1.7140 - val_accuracy: 0.6351\n",
      "Epoch 50/100\n",
      "143934/143934 [==============================] - 84s 580us/sample - loss: 0.4902 - accuracy: 0.8431 - val_loss: 1.4501 - val_accuracy: 0.6126\n",
      "Evaluating model...\n",
      "ROC: 0.952\n",
      "\n",
      "Accuracy = 0.68\n",
      "F-Score: 0.68\n",
      "\n",
      "*** Train on {1, 3, 4, 6, 7, 8, 9} Validate on 5 Test on 2 ***\n",
      "val shape:  (936, 128, 128, 1)\n",
      "test shape:  (888, 128, 128, 1)\n",
      "train shape:  (145068, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 145068 samples, validate on 888 samples\n",
      "Epoch 1/100\n",
      "145068/145068 [==============================] - 86s 592us/sample - loss: 2.1636 - accuracy: 0.2599 - val_loss: 1.7215 - val_accuracy: 0.4707\n",
      "Epoch 2/100\n",
      "145068/145068 [==============================] - 84s 578us/sample - loss: 1.6935 - accuracy: 0.4479 - val_loss: 1.3639 - val_accuracy: 0.5541\n",
      "Epoch 3/100\n",
      "145068/145068 [==============================] - 84s 577us/sample - loss: 1.4734 - accuracy: 0.5304 - val_loss: 1.2175 - val_accuracy: 0.5991\n",
      "Epoch 4/100\n",
      "145068/145068 [==============================] - 83s 575us/sample - loss: 1.3381 - accuracy: 0.5775 - val_loss: 1.2775 - val_accuracy: 0.5642\n",
      "Epoch 5/100\n",
      "145068/145068 [==============================] - 84s 577us/sample - loss: 1.2485 - accuracy: 0.6066 - val_loss: 1.3359 - val_accuracy: 0.5541\n",
      "Epoch 6/100\n",
      "145068/145068 [==============================] - 84s 577us/sample - loss: 1.1735 - accuracy: 0.6308 - val_loss: 1.2721 - val_accuracy: 0.5518\n",
      "Epoch 7/100\n",
      "145068/145068 [==============================] - 84s 577us/sample - loss: 1.1094 - accuracy: 0.6512 - val_loss: 1.1726 - val_accuracy: 0.6171\n",
      "Epoch 8/100\n",
      "145068/145068 [==============================] - 84s 577us/sample - loss: 1.0620 - accuracy: 0.6674 - val_loss: 1.1925 - val_accuracy: 0.5923\n",
      "Epoch 9/100\n",
      "145068/145068 [==============================] - 84s 578us/sample - loss: 1.0141 - accuracy: 0.6822 - val_loss: 1.1733 - val_accuracy: 0.6104\n",
      "Epoch 10/100\n",
      "145068/145068 [==============================] - 84s 579us/sample - loss: 0.9832 - accuracy: 0.6922 - val_loss: 1.2813 - val_accuracy: 0.5946\n",
      "Epoch 11/100\n",
      "145068/145068 [==============================] - 84s 579us/sample - loss: 0.9453 - accuracy: 0.7029 - val_loss: 1.1558 - val_accuracy: 0.6408\n",
      "Epoch 12/100\n",
      "145068/145068 [==============================] - 84s 581us/sample - loss: 0.9174 - accuracy: 0.7120 - val_loss: 1.3075 - val_accuracy: 0.5822\n",
      "Epoch 13/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.8906 - accuracy: 0.7189 - val_loss: 1.3633 - val_accuracy: 0.5721\n",
      "Epoch 14/100\n",
      "145068/145068 [==============================] - 84s 578us/sample - loss: 0.8710 - accuracy: 0.7249 - val_loss: 1.1795 - val_accuracy: 0.6273\n",
      "Epoch 15/100\n",
      "145068/145068 [==============================] - 84s 582us/sample - loss: 0.8501 - accuracy: 0.7310 - val_loss: 1.0965 - val_accuracy: 0.6374\n",
      "Epoch 16/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.8236 - accuracy: 0.7392 - val_loss: 1.7861 - val_accuracy: 0.5135\n",
      "Epoch 17/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.8095 - accuracy: 0.7431 - val_loss: 1.5869 - val_accuracy: 0.5360\n",
      "Epoch 18/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.7938 - accuracy: 0.7478 - val_loss: 1.1426 - val_accuracy: 0.6374\n",
      "Epoch 19/100\n",
      "145068/145068 [==============================] - 85s 584us/sample - loss: 0.7776 - accuracy: 0.7523 - val_loss: 1.1594 - val_accuracy: 0.6385\n",
      "Epoch 20/100\n",
      "145068/145068 [==============================] - 85s 584us/sample - loss: 0.7654 - accuracy: 0.7574 - val_loss: 1.3865 - val_accuracy: 0.5788\n",
      "Epoch 21/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.7493 - accuracy: 0.7622 - val_loss: 1.2540 - val_accuracy: 0.6227\n",
      "Epoch 22/100\n",
      "145068/145068 [==============================] - 85s 584us/sample - loss: 0.7380 - accuracy: 0.7647 - val_loss: 1.2066 - val_accuracy: 0.6351\n",
      "Epoch 23/100\n",
      "145068/145068 [==============================] - 84s 582us/sample - loss: 0.7296 - accuracy: 0.7676 - val_loss: 1.2399 - val_accuracy: 0.6273\n",
      "Epoch 24/100\n",
      "145068/145068 [==============================] - 85s 584us/sample - loss: 0.7164 - accuracy: 0.7725 - val_loss: 1.4128 - val_accuracy: 0.5867\n",
      "Epoch 25/100\n",
      "145068/145068 [==============================] - 85s 584us/sample - loss: 0.7103 - accuracy: 0.7735 - val_loss: 1.2663 - val_accuracy: 0.6306\n",
      "Epoch 26/100\n",
      "145068/145068 [==============================] - 85s 585us/sample - loss: 0.6972 - accuracy: 0.7782 - val_loss: 1.2256 - val_accuracy: 0.6385\n",
      "Epoch 27/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.6899 - accuracy: 0.7807 - val_loss: 1.2468 - val_accuracy: 0.6104\n",
      "Epoch 28/100\n",
      "145068/145068 [==============================] - 85s 585us/sample - loss: 0.6797 - accuracy: 0.7841 - val_loss: 1.3269 - val_accuracy: 0.6250\n",
      "Epoch 29/100\n",
      "145068/145068 [==============================] - 85s 587us/sample - loss: 0.6759 - accuracy: 0.7850 - val_loss: 1.5040 - val_accuracy: 0.5901\n",
      "Epoch 30/100\n",
      "145068/145068 [==============================] - 85s 583us/sample - loss: 0.6704 - accuracy: 0.7863 - val_loss: 1.3651 - val_accuracy: 0.6092\n",
      "Evaluating model...\n",
      "ROC: 0.955\n",
      "\n",
      "Accuracy = 0.73\n",
      "F-Score: 0.73\n",
      "\n",
      "*** Train on {1, 3, 4, 5, 7, 8, 9} Validate on 6 Test on 2 ***\n",
      "val shape:  (823, 128, 128, 1)\n",
      "test shape:  (888, 128, 128, 1)\n",
      "train shape:  (147441, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 147441 samples, validate on 888 samples\n",
      "Epoch 1/100\n",
      "147441/147441 [==============================] - 84s 569us/sample - loss: 2.1739 - accuracy: 0.2549 - val_loss: 1.6863 - val_accuracy: 0.4673\n",
      "Epoch 2/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 1.6909 - accuracy: 0.4472 - val_loss: 1.4096 - val_accuracy: 0.4932\n",
      "Epoch 3/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 1.4607 - accuracy: 0.5301 - val_loss: 1.5925 - val_accuracy: 0.4550\n",
      "Epoch 4/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 1.3297 - accuracy: 0.5756 - val_loss: 1.3217 - val_accuracy: 0.5541 accu\n",
      "Epoch 5/100\n",
      "147441/147441 [==============================] - 86s 583us/sample - loss: 1.2320 - accuracy: 0.6107 - val_loss: 1.2371 - val_accuracy: 0.5845\n",
      "Epoch 6/100\n",
      "147441/147441 [==============================] - 86s 583us/sample - loss: 1.1507 - accuracy: 0.6375 - val_loss: 1.2861 - val_accuracy: 0.5462\n",
      "Epoch 7/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "147441/147441 [==============================] - 86s 582us/sample - loss: 1.0928 - accuracy: 0.6562 - val_loss: 1.1640 - val_accuracy: 0.6273\n",
      "Epoch 8/100\n",
      "147441/147441 [==============================] - 86s 584us/sample - loss: 1.0382 - accuracy: 0.6743 - val_loss: 1.3111 - val_accuracy: 0.5957\n",
      "Epoch 9/100\n",
      "147441/147441 [==============================] - 86s 586us/sample - loss: 0.9917 - accuracy: 0.6903 - val_loss: 1.5384 - val_accuracy: 0.5608\n",
      "Epoch 10/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.9490 - accuracy: 0.7017 - val_loss: 1.1756 - val_accuracy: 0.6126\n",
      "Epoch 11/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.9124 - accuracy: 0.7131 - val_loss: 1.3590 - val_accuracy: 0.5822\n",
      "Epoch 12/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.8826 - accuracy: 0.7232 - val_loss: 1.1411 - val_accuracy: 0.6374\n",
      "Epoch 13/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.8476 - accuracy: 0.7351 - val_loss: 1.1314 - val_accuracy: 0.6284\n",
      "Epoch 14/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.8232 - accuracy: 0.7417 - val_loss: 1.1840 - val_accuracy: 0.6374\n",
      "Epoch 15/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.7980 - accuracy: 0.7503 - val_loss: 1.1949 - val_accuracy: 0.6216\n",
      "Epoch 16/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.7725 - accuracy: 0.7572 - val_loss: 1.6242 - val_accuracy: 0.5191\n",
      "Epoch 17/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.7556 - accuracy: 0.7627 - val_loss: 1.2078 - val_accuracy: 0.6014\n",
      "Epoch 18/100\n",
      "147441/147441 [==============================] - 86s 580us/sample - loss: 0.7573 - accuracy: 0.7620 - val_loss: 1.0838 - val_accuracy: 0.6610\n",
      "Epoch 19/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.7215 - accuracy: 0.7737 - val_loss: 1.1839 - val_accuracy: 0.6261\n",
      "Epoch 20/100\n",
      "147441/147441 [==============================] - 86s 580us/sample - loss: 0.7052 - accuracy: 0.7789 - val_loss: 1.3048 - val_accuracy: 0.6363\n",
      "Epoch 21/100\n",
      "147441/147441 [==============================] - 85s 578us/sample - loss: 0.6886 - accuracy: 0.7833 - val_loss: 1.2062 - val_accuracy: 0.6104\n",
      "Epoch 22/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.6791 - accuracy: 0.7854 - val_loss: 1.1297 - val_accuracy: 0.6577\n",
      "Epoch 23/100\n",
      "147441/147441 [==============================] - 85s 578us/sample - loss: 0.6671 - accuracy: 0.7904 - val_loss: 1.1941 - val_accuracy: 0.6126\n",
      "Epoch 24/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.6503 - accuracy: 0.7945 - val_loss: 1.2574 - val_accuracy: 0.6216\n",
      "Epoch 25/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.6321 - accuracy: 0.8004 - val_loss: 1.2027 - val_accuracy: 0.6182\n",
      "Epoch 26/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.6283 - accuracy: 0.8021 - val_loss: 1.1821 - val_accuracy: 0.6532\n",
      "Epoch 27/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.6098 - accuracy: 0.8078 - val_loss: 1.1922 - val_accuracy: 0.6430\n",
      "Epoch 28/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.5992 - accuracy: 0.8108 - val_loss: 1.1096 - val_accuracy: 0.6520\n",
      "Epoch 29/100\n",
      "147441/147441 [==============================] - 86s 585us/sample - loss: 0.5854 - accuracy: 0.8162 - val_loss: 1.1747 - val_accuracy: 0.6430\n",
      "Epoch 30/100\n",
      "147441/147441 [==============================] - 86s 586us/sample - loss: 0.5779 - accuracy: 0.8190 - val_loss: 1.5663 - val_accuracy: 0.6329\n",
      "Epoch 31/100\n",
      "147441/147441 [==============================] - 86s 583us/sample - loss: 0.5677 - accuracy: 0.8209 - val_loss: 1.0763 - val_accuracy: 0.6453\n",
      "Epoch 32/100\n",
      "147441/147441 [==============================] - 86s 583us/sample - loss: 0.5541 - accuracy: 0.8244 - val_loss: 1.2573 - val_accuracy: 0.5935\n",
      "Epoch 33/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.5432 - accuracy: 0.8287 - val_loss: 1.0755 - val_accuracy: 0.6700\n",
      "Epoch 34/100\n",
      "147441/147441 [==============================] - 86s 584us/sample - loss: 0.5354 - accuracy: 0.8314 - val_loss: 1.3054 - val_accuracy: 0.6408\n",
      "Epoch 35/100\n",
      "147441/147441 [==============================] - 86s 584us/sample - loss: 0.5259 - accuracy: 0.8347 - val_loss: 1.1704 - val_accuracy: 0.6588\n",
      "Epoch 36/100\n",
      "147441/147441 [==============================] - 87s 592us/sample - loss: 0.5185 - accuracy: 0.8366 - val_loss: 1.0152 - val_accuracy: 0.6745\n",
      "Epoch 37/100\n",
      "147441/147441 [==============================] - 89s 601us/sample - loss: 0.5113 - accuracy: 0.8390 - val_loss: 1.2132 - val_accuracy: 0.6498\n",
      "Epoch 38/100\n",
      "147441/147441 [==============================] - 87s 590us/sample - loss: 0.5069 - accuracy: 0.8414 - val_loss: 1.1012 - val_accuracy: 0.6678\n",
      "Epoch 39/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4975 - accuracy: 0.8431 - val_loss: 1.0851 - val_accuracy: 0.6453\n",
      "Epoch 40/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.4883 - accuracy: 0.8473 - val_loss: 1.1668 - val_accuracy: 0.6396\n",
      "Epoch 41/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4871 - accuracy: 0.8467 - val_loss: 1.2125 - val_accuracy: 0.6498\n",
      "Epoch 42/100\n",
      "147441/147441 [==============================] - 86s 583us/sample - loss: 0.4762 - accuracy: 0.8507 - val_loss: 1.1815 - val_accuracy: 0.6396\n",
      "Epoch 43/100\n",
      "147441/147441 [==============================] - 86s 583us/sample - loss: 0.4734 - accuracy: 0.8510 - val_loss: 1.1737 - val_accuracy: 0.6543\n",
      "Epoch 44/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4652 - accuracy: 0.8537 - val_loss: 1.1688 - val_accuracy: 0.6565\n",
      "Epoch 45/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4609 - accuracy: 0.8554 - val_loss: 0.9354 - val_accuracy: 0.7140\n",
      "Epoch 46/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4564 - accuracy: 0.8570 - val_loss: 1.1623 - val_accuracy: 0.6577\n",
      "Epoch 47/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4489 - accuracy: 0.8585 - val_loss: 1.2646 - val_accuracy: 0.6779\n",
      "Epoch 48/100\n",
      "147441/147441 [==============================] - 86s 582us/sample - loss: 0.4498 - accuracy: 0.8595 - val_loss: 1.0820 - val_accuracy: 0.6869\n",
      "Epoch 49/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.4427 - accuracy: 0.8615 - val_loss: 1.3721 - val_accuracy: 0.6475\n",
      "Epoch 50/100\n",
      "147441/147441 [==============================] - 86s 581us/sample - loss: 0.4427 - accuracy: 0.8612 - val_loss: 1.0345 - val_accuracy: 0.6678\n",
      "Epoch 51/100\n",
      "147441/147441 [==============================] - 86s 580us/sample - loss: 0.4348 - accuracy: 0.8639 - val_loss: 1.1906 - val_accuracy: 0.6712\n",
      "Epoch 52/100\n",
      "147441/147441 [==============================] - 86s 580us/sample - loss: 0.4302 - accuracy: 0.8653 - val_loss: 1.0676 - val_accuracy: 0.6982\n",
      "Epoch 53/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.4265 - accuracy: 0.8668 - val_loss: 1.2464 - val_accuracy: 0.6655\n",
      "Epoch 54/100\n",
      "147441/147441 [==============================] - 85s 580us/sample - loss: 0.4212 - accuracy: 0.8677 - val_loss: 1.0772 - val_accuracy: 0.7117\n",
      "Epoch 55/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.4174 - accuracy: 0.8697 - val_loss: 1.3874 - val_accuracy: 0.6948\n",
      "Epoch 56/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.4171 - accuracy: 0.8699 - val_loss: 1.2108 - val_accuracy: 0.6914\n",
      "Epoch 57/100\n",
      "147441/147441 [==============================] - 85s 580us/sample - loss: 0.4098 - accuracy: 0.8718 - val_loss: 1.1880 - val_accuracy: 0.6959\n",
      "Epoch 58/100\n",
      "147441/147441 [==============================] - 85s 579us/sample - loss: 0.4063 - accuracy: 0.8734 - val_loss: 1.1669 - val_accuracy: 0.6892\n",
      "Epoch 59/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "147441/147441 [==============================] - 85s 577us/sample - loss: 0.4138 - accuracy: 0.8708 - val_loss: 1.2678 - val_accuracy: 0.6745\n",
      "Epoch 60/100\n",
      "147441/147441 [==============================] - 85s 577us/sample - loss: 0.4017 - accuracy: 0.8742 - val_loss: 1.1637 - val_accuracy: 0.6937\n",
      "Evaluating model...\n",
      "ROC: 0.955\n",
      "\n",
      "Accuracy = 0.74\n",
      "F-Score: 0.74\n",
      "\n",
      "*** Train on {1, 2, 4, 5, 6, 8, 9} Validate on 7 Test on 3 ***\n",
      "val shape:  (838, 128, 128, 1)\n",
      "test shape:  (925, 128, 128, 1)\n",
      "train shape:  (146349, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 146349 samples, validate on 925 samples\n",
      "Epoch 1/100\n",
      "146349/146349 [==============================] - 87s 593us/sample - loss: 2.5131 - accuracy: 0.1408 - val_loss: 2.1388 - val_accuracy: 0.3005\n",
      "Epoch 2/100\n",
      "146349/146349 [==============================] - 84s 573us/sample - loss: 1.9021 - accuracy: 0.3683 - val_loss: 1.7868 - val_accuracy: 0.3773\n",
      "Epoch 3/100\n",
      "146349/146349 [==============================] - 84s 575us/sample - loss: 1.5745 - accuracy: 0.4907 - val_loss: 1.6138 - val_accuracy: 0.4195\n",
      "Epoch 4/100\n",
      "146349/146349 [==============================] - 84s 574us/sample - loss: 1.4091 - accuracy: 0.5541 - val_loss: 1.5880 - val_accuracy: 0.4659\n",
      "Epoch 5/100\n",
      "146349/146349 [==============================] - 84s 575us/sample - loss: 1.2976 - accuracy: 0.5950 - val_loss: 1.5284 - val_accuracy: 0.4649\n",
      "Epoch 6/100\n",
      "146349/146349 [==============================] - 84s 575us/sample - loss: 1.2041 - accuracy: 0.6266 - val_loss: 1.4631 - val_accuracy: 0.4876\n",
      "Epoch 7/100\n",
      "146349/146349 [==============================] - 84s 575us/sample - loss: 1.1390 - accuracy: 0.6488 - val_loss: 1.5695 - val_accuracy: 0.4886\n",
      "Epoch 8/100\n",
      "146349/146349 [==============================] - 85s 578us/sample - loss: 1.0786 - accuracy: 0.6673 - val_loss: 1.3281 - val_accuracy: 0.5470\n",
      "Epoch 9/100\n",
      "146349/146349 [==============================] - 84s 576us/sample - loss: 1.0361 - accuracy: 0.6796 - val_loss: 1.6495 - val_accuracy: 0.4757\n",
      "Epoch 10/100\n",
      "146349/146349 [==============================] - 84s 577us/sample - loss: 0.9952 - accuracy: 0.6941 - val_loss: 1.4946 - val_accuracy: 0.5103\n",
      "Epoch 11/100\n",
      "146349/146349 [==============================] - 84s 576us/sample - loss: 0.9591 - accuracy: 0.7035 - val_loss: 1.6760 - val_accuracy: 0.4951\n",
      "Epoch 12/100\n",
      "146349/146349 [==============================] - 84s 574us/sample - loss: 0.9316 - accuracy: 0.7119 - val_loss: 1.3803 - val_accuracy: 0.5330\n",
      "Epoch 13/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.8993 - accuracy: 0.7210 - val_loss: 1.4436 - val_accuracy: 0.5470\n",
      "Epoch 14/100\n",
      "146349/146349 [==============================] - 87s 593us/sample - loss: 0.8745 - accuracy: 0.7280 - val_loss: 1.3690 - val_accuracy: 0.5589\n",
      "Epoch 15/100\n",
      "146349/146349 [==============================] - 86s 590us/sample - loss: 0.8514 - accuracy: 0.7361 - val_loss: 1.4504 - val_accuracy: 0.5524\n",
      "Epoch 16/100\n",
      "146349/146349 [==============================] - 86s 590us/sample - loss: 0.8286 - accuracy: 0.7414 - val_loss: 1.4138 - val_accuracy: 0.5503\n",
      "Epoch 17/100\n",
      "146349/146349 [==============================] - 87s 592us/sample - loss: 0.8109 - accuracy: 0.7469 - val_loss: 1.3898 - val_accuracy: 0.5589\n",
      "Epoch 18/100\n",
      "146349/146349 [==============================] - 86s 591us/sample - loss: 0.7918 - accuracy: 0.7522 - val_loss: 1.3508 - val_accuracy: 0.5730\n",
      "Epoch 19/100\n",
      "146349/146349 [==============================] - 86s 588us/sample - loss: 0.7767 - accuracy: 0.7574 - val_loss: 1.3755 - val_accuracy: 0.5535\n",
      "Epoch 20/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.7603 - accuracy: 0.7624 - val_loss: 1.6541 - val_accuracy: 0.4908\n",
      "Epoch 21/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.7475 - accuracy: 0.7649 - val_loss: 1.3331 - val_accuracy: 0.5600\n",
      "Epoch 22/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.7377 - accuracy: 0.7681 - val_loss: 1.4429 - val_accuracy: 0.5622\n",
      "Epoch 23/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.7264 - accuracy: 0.7695 - val_loss: 1.3200 - val_accuracy: 0.5762\n",
      "Epoch 24/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.7114 - accuracy: 0.7757 - val_loss: 1.7466 - val_accuracy: 0.4962\n",
      "Epoch 25/100\n",
      "146349/146349 [==============================] - 85s 582us/sample - loss: 0.7026 - accuracy: 0.7776 - val_loss: 1.3991 - val_accuracy: 0.5654\n",
      "Epoch 26/100\n",
      "146349/146349 [==============================] - 85s 580us/sample - loss: 0.6922 - accuracy: 0.7806 - val_loss: 1.4652 - val_accuracy: 0.5524\n",
      "Epoch 27/100\n",
      "146349/146349 [==============================] - 84s 576us/sample - loss: 0.6831 - accuracy: 0.7833 - val_loss: 1.3770 - val_accuracy: 0.5719\n",
      "Epoch 28/100\n",
      "146349/146349 [==============================] - 85s 578us/sample - loss: 0.6682 - accuracy: 0.7874 - val_loss: 1.3432 - val_accuracy: 0.5784\n",
      "Epoch 29/100\n",
      "146349/146349 [==============================] - 85s 579us/sample - loss: 0.6607 - accuracy: 0.7902 - val_loss: 1.3206 - val_accuracy: 0.5719\n",
      "Epoch 30/100\n",
      "146349/146349 [==============================] - 85s 581us/sample - loss: 0.6590 - accuracy: 0.7912 - val_loss: 1.5034 - val_accuracy: 0.5492\n",
      "Epoch 31/100\n",
      "146349/146349 [==============================] - 86s 587us/sample - loss: 0.6495 - accuracy: 0.7915 - val_loss: 1.3196 - val_accuracy: 0.5784\n",
      "Epoch 32/100\n",
      "146349/146349 [==============================] - 86s 587us/sample - loss: 0.6457 - accuracy: 0.7939 - val_loss: 1.3170 - val_accuracy: 0.5935\n",
      "Epoch 33/100\n",
      "146349/146349 [==============================] - 85s 584us/sample - loss: 0.6355 - accuracy: 0.7965 - val_loss: 1.4644 - val_accuracy: 0.5730\n",
      "Epoch 34/100\n",
      "146349/146349 [==============================] - 86s 585us/sample - loss: 0.6322 - accuracy: 0.7984 - val_loss: 1.3829 - val_accuracy: 0.5795\n",
      "Epoch 35/100\n",
      "146349/146349 [==============================] - 85s 584us/sample - loss: 0.6231 - accuracy: 0.8014 - val_loss: 1.3254 - val_accuracy: 0.5816\n",
      "Epoch 36/100\n",
      "146349/146349 [==============================] - 85s 583us/sample - loss: 0.6170 - accuracy: 0.8021 - val_loss: 1.4611 - val_accuracy: 0.5622\n",
      "Epoch 37/100\n",
      "146349/146349 [==============================] - 85s 583us/sample - loss: 0.6133 - accuracy: 0.8042 - val_loss: 1.4112 - val_accuracy: 0.5805\n",
      "Epoch 38/100\n",
      "146349/146349 [==============================] - 84s 576us/sample - loss: 0.6007 - accuracy: 0.8094 - val_loss: 1.4080 - val_accuracy: 0.5849\n",
      "Epoch 39/100\n",
      "146349/146349 [==============================] - 86s 588us/sample - loss: 0.5980 - accuracy: 0.8082 - val_loss: 1.3839 - val_accuracy: 0.5665\n",
      "Epoch 40/100\n",
      "146349/146349 [==============================] - 86s 587us/sample - loss: 0.5946 - accuracy: 0.8094 - val_loss: 1.4374 - val_accuracy: 0.5481\n",
      "Epoch 41/100\n",
      "146349/146349 [==============================] - 86s 586us/sample - loss: 0.5863 - accuracy: 0.8124 - val_loss: 1.5686 - val_accuracy: 0.5827\n",
      "Epoch 42/100\n",
      "146349/146349 [==============================] - 85s 583us/sample - loss: 0.5837 - accuracy: 0.8136 - val_loss: 1.6728 - val_accuracy: 0.5568\n",
      "Epoch 43/100\n",
      "146349/146349 [==============================] - 85s 583us/sample - loss: 0.5799 - accuracy: 0.8147 - val_loss: 1.4820 - val_accuracy: 0.5470\n",
      "Epoch 44/100\n",
      "146349/146349 [==============================] - 86s 585us/sample - loss: 0.5763 - accuracy: 0.8153 - val_loss: 1.3463 - val_accuracy: 0.5957\n",
      "Epoch 45/100\n",
      "146349/146349 [==============================] - 86s 585us/sample - loss: 0.5716 - accuracy: 0.8164 - val_loss: 1.4994 - val_accuracy: 0.5622\n",
      "Epoch 46/100\n",
      "146349/146349 [==============================] - 85s 582us/sample - loss: 0.5688 - accuracy: 0.8188 - val_loss: 1.4142 - val_accuracy: 0.5859\n",
      "Epoch 47/100\n",
      "146349/146349 [==============================] - 85s 583us/sample - loss: 0.5630 - accuracy: 0.8200 - val_loss: 1.3644 - val_accuracy: 0.5978\n",
      "Evaluating model...\n",
      "ROC: 0.954\n",
      "\n",
      "Accuracy = 0.65\n",
      "F-Score: 0.65\n",
      "\n",
      "*** Train on {1, 2, 3, 4, 5, 6, 9} Validate on 8 Test on 7 ***\n",
      "val shape:  (806, 128, 128, 1)\n",
      "test shape:  (838, 128, 128, 1)\n",
      "train shape:  (148848, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 148848 samples, validate on 838 samples\n",
      "Epoch 1/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "148848/148848 [==============================] - 87s 586us/sample - loss: 2.1218 - accuracy: 0.2796 - val_loss: 1.5991 - val_accuracy: 0.4976\n",
      "Epoch 2/100\n",
      "148848/148848 [==============================] - 87s 584us/sample - loss: 1.4353 - accuracy: 0.5410 - val_loss: 1.2616 - val_accuracy: 0.6265\n",
      "Epoch 3/100\n",
      "148848/148848 [==============================] - 87s 583us/sample - loss: 1.1803 - accuracy: 0.6367 - val_loss: 1.1315 - val_accuracy: 0.6659\n",
      "Epoch 4/100\n",
      "148848/148848 [==============================] - 87s 586us/sample - loss: 1.0423 - accuracy: 0.6835 - val_loss: 1.1274 - val_accuracy: 0.6492\n",
      "Epoch 5/100\n",
      "148848/148848 [==============================] - 87s 586us/sample - loss: 0.9774 - accuracy: 0.7034 - val_loss: 1.0139 - val_accuracy: 0.6933\n",
      "Epoch 6/100\n",
      "148848/148848 [==============================] - 87s 584us/sample - loss: 0.8838 - accuracy: 0.7352 - val_loss: 0.9898 - val_accuracy: 0.6957\n",
      "Epoch 7/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.8235 - accuracy: 0.7543 - val_loss: 1.1195 - val_accuracy: 0.6778\n",
      "Epoch 8/100\n",
      "148848/148848 [==============================] - 87s 584us/sample - loss: 0.7796 - accuracy: 0.7681 - val_loss: 1.1586 - val_accuracy: 0.6587\n",
      "Epoch 9/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.7476 - accuracy: 0.7772 - val_loss: 1.0299 - val_accuracy: 0.6909\n",
      "Epoch 10/100\n",
      "148848/148848 [==============================] - 87s 583us/sample - loss: 0.7136 - accuracy: 0.7873 - val_loss: 1.0543 - val_accuracy: 0.7124\n",
      "Epoch 11/100\n",
      "148848/148848 [==============================] - 87s 582us/sample - loss: 0.6879 - accuracy: 0.7953 - val_loss: 1.0711 - val_accuracy: 0.7005\n",
      "Epoch 12/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.6616 - accuracy: 0.8023 - val_loss: 1.2009 - val_accuracy: 0.6802\n",
      "Epoch 13/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.6413 - accuracy: 0.8082 - val_loss: 1.2530 - val_accuracy: 0.6599\n",
      "Epoch 14/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.6198 - accuracy: 0.8139 - val_loss: 1.0928 - val_accuracy: 0.7029\n",
      "Epoch 15/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.5994 - accuracy: 0.8208 - val_loss: 1.1414 - val_accuracy: 0.6945\n",
      "Epoch 16/100\n",
      "148848/148848 [==============================] - 87s 584us/sample - loss: 0.5888 - accuracy: 0.8234 - val_loss: 1.1047 - val_accuracy: 0.7005\n",
      "Epoch 17/100\n",
      "148848/148848 [==============================] - 87s 584us/sample - loss: 0.5726 - accuracy: 0.8277 - val_loss: 0.9970 - val_accuracy: 0.7399\n",
      "Epoch 18/100\n",
      "148848/148848 [==============================] - 87s 584us/sample - loss: 0.5623 - accuracy: 0.8298 - val_loss: 1.1479 - val_accuracy: 0.7017\n",
      "Epoch 19/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.5450 - accuracy: 0.8346 - val_loss: 1.0925 - val_accuracy: 0.7172\n",
      "Epoch 20/100\n",
      "148848/148848 [==============================] - 87s 585us/sample - loss: 0.5305 - accuracy: 0.8398 - val_loss: 1.0745 - val_accuracy: 0.7267\n",
      "Epoch 21/100\n",
      "148848/148848 [==============================] - 87s 583us/sample - loss: 0.5235 - accuracy: 0.8417 - val_loss: 1.1594 - val_accuracy: 0.7053\n",
      "Evaluating model...\n",
      "ROC: 0.904\n",
      "\n",
      "Accuracy = 0.59\n",
      "F-Score: 0.59\n",
      "\n",
      "*** Train on {1, 2, 4, 5, 6, 7, 8} Validate on 9 Test on 3 ***\n",
      "val shape:  (816, 128, 128, 1)\n",
      "test shape:  (925, 128, 128, 1)\n",
      "train shape:  (146811, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 146811 samples, validate on 925 samples\n",
      "Epoch 1/100\n",
      "146811/146811 [==============================] - 82s 561us/sample - loss: 2.4057 - accuracy: 0.1911 - val_loss: 1.9467 - val_accuracy: 0.3319\n",
      "Epoch 2/100\n",
      "146811/146811 [==============================] - 83s 565us/sample - loss: 1.7569 - accuracy: 0.4227 - val_loss: 1.7025 - val_accuracy: 0.4195\n",
      "Epoch 3/100\n",
      "146811/146811 [==============================] - 83s 566us/sample - loss: 1.5122 - accuracy: 0.5164 - val_loss: 2.3725 - val_accuracy: 0.3427\n",
      "Epoch 4/100\n",
      "146811/146811 [==============================] - 83s 566us/sample - loss: 1.3629 - accuracy: 0.5730 - val_loss: 1.5641 - val_accuracy: 0.4876\n",
      "Epoch 5/100\n",
      "146811/146811 [==============================] - 84s 570us/sample - loss: 1.2504 - accuracy: 0.6147 - val_loss: 1.5914 - val_accuracy: 0.4724\n",
      "Epoch 6/100\n",
      "146811/146811 [==============================] - 83s 565us/sample - loss: 1.1640 - accuracy: 0.6444 - val_loss: 3.3988 - val_accuracy: 0.2389\n",
      "Epoch 7/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 1.1566 - accuracy: 0.6445 - val_loss: 1.5085 - val_accuracy: 0.5038\n",
      "Epoch 8/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 1.0996 - accuracy: 0.6619 - val_loss: 1.3331 - val_accuracy: 0.5438\n",
      "Epoch 9/100\n",
      "146811/146811 [==============================] - 83s 568us/sample - loss: 1.0536 - accuracy: 0.6759 - val_loss: 1.4805 - val_accuracy: 0.4832\n",
      "Epoch 10/100\n",
      "146811/146811 [==============================] - 83s 566us/sample - loss: 1.0163 - accuracy: 0.6876 - val_loss: 1.3551 - val_accuracy: 0.5146\n",
      "Epoch 11/100\n",
      "146811/146811 [==============================] - 83s 565us/sample - loss: 0.9834 - accuracy: 0.6975 - val_loss: 1.5851 - val_accuracy: 0.4681\n",
      "Epoch 12/100\n",
      "146811/146811 [==============================] - 83s 568us/sample - loss: 0.9547 - accuracy: 0.7050 - val_loss: 1.6163 - val_accuracy: 0.4703\n",
      "Epoch 13/100\n",
      "146811/146811 [==============================] - 84s 573us/sample - loss: 0.9333 - accuracy: 0.7126 - val_loss: 1.3632 - val_accuracy: 0.5643\n",
      "Epoch 14/100\n",
      "146811/146811 [==============================] - 84s 571us/sample - loss: 0.9100 - accuracy: 0.7176 - val_loss: 1.5368 - val_accuracy: 0.5297\n",
      "Epoch 15/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 0.8892 - accuracy: 0.7237 - val_loss: 1.3484 - val_accuracy: 0.5773\n",
      "Epoch 16/100\n",
      "146811/146811 [==============================] - 84s 572us/sample - loss: 0.8641 - accuracy: 0.7318 - val_loss: 1.5906 - val_accuracy: 0.5103\n",
      "Epoch 17/100\n",
      "146811/146811 [==============================] - 84s 571us/sample - loss: 0.8505 - accuracy: 0.7336 - val_loss: 1.3349 - val_accuracy: 0.5643\n",
      "Epoch 18/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 0.8313 - accuracy: 0.7413 - val_loss: 1.3064 - val_accuracy: 0.5524\n",
      "Epoch 19/100\n",
      "146811/146811 [==============================] - 84s 571us/sample - loss: 0.8166 - accuracy: 0.7429 - val_loss: 1.6962 - val_accuracy: 0.5265\n",
      "Epoch 20/100\n",
      "146811/146811 [==============================] - 84s 571us/sample - loss: 0.8035 - accuracy: 0.7471 - val_loss: 1.4264 - val_accuracy: 0.5535\n",
      "Epoch 21/100\n",
      "146811/146811 [==============================] - 83s 568us/sample - loss: 0.7876 - accuracy: 0.7521 - val_loss: 1.3323 - val_accuracy: 0.5665\n",
      "Epoch 22/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 0.7656 - accuracy: 0.7587 - val_loss: 1.4201 - val_accuracy: 0.5643\n",
      "Epoch 23/100\n",
      "146811/146811 [==============================] - 83s 567us/sample - loss: 0.7534 - accuracy: 0.7627 - val_loss: 1.4005 - val_accuracy: 0.5557\n",
      "Epoch 24/100\n",
      "146811/146811 [==============================] - 83s 567us/sample - loss: 0.7539 - accuracy: 0.7611 - val_loss: 1.3137 - val_accuracy: 0.5578\n",
      "Epoch 25/100\n",
      "146811/146811 [==============================] - 83s 568us/sample - loss: 0.7434 - accuracy: 0.7655 - val_loss: 1.6499 - val_accuracy: 0.5254\n",
      "Epoch 26/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 0.7345 - accuracy: 0.7671 - val_loss: 1.6892 - val_accuracy: 0.4886\n",
      "Epoch 27/100\n",
      "146811/146811 [==============================] - 83s 567us/sample - loss: 0.7259 - accuracy: 0.7701 - val_loss: 1.3784 - val_accuracy: 0.5524\n",
      "Epoch 28/100\n",
      "146811/146811 [==============================] - 84s 570us/sample - loss: 0.7159 - accuracy: 0.7712 - val_loss: 1.8677 - val_accuracy: 0.4659\n",
      "Epoch 29/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "146811/146811 [==============================] - 84s 574us/sample - loss: 0.7045 - accuracy: 0.7752 - val_loss: 1.9428 - val_accuracy: 0.4789045 - accuracy: 0.77\n",
      "Epoch 30/100\n",
      "146811/146811 [==============================] - 84s 570us/sample - loss: 0.6815 - accuracy: 0.7818 - val_loss: 1.5044 - val_accuracy: 0.5557\n",
      "Epoch 31/100\n",
      "146811/146811 [==============================] - 84s 571us/sample - loss: 0.6693 - accuracy: 0.7868 - val_loss: 1.8426 - val_accuracy: 0.5178\n",
      "Epoch 32/100\n",
      "146811/146811 [==============================] - 84s 571us/sample - loss: 0.6644 - accuracy: 0.7883 - val_loss: 1.4606 - val_accuracy: 0.5741\n",
      "Epoch 33/100\n",
      "146811/146811 [==============================] - 84s 569us/sample - loss: 0.6559 - accuracy: 0.7917 - val_loss: 1.5521 - val_accuracy: 0.5730\n",
      "Evaluating model...\n",
      "ROC: 0.932\n",
      "\n",
      "Accuracy = 0.65\n",
      "F-Score: 0.65\n",
      "\n",
      "*** Train on {1, 2, 3, 4, 5, 7, 8, 9} Validate on 10 Test on 6 ***\n",
      "val shape:  (837, 128, 128, 1)\n",
      "test shape:  (823, 128, 128, 1)\n",
      "train shape:  (148512, 128, 128, 1)\n",
      "Building model...\n",
      "Training model...\n",
      "Train on 148512 samples, validate on 823 samples\n",
      "Epoch 1/100\n",
      "148512/148512 [==============================] - 89s 602us/sample - loss: 2.0152 - accuracy: 0.3237 - val_loss: 2.1258 - val_accuracy: 0.3317\n",
      "Epoch 2/100\n",
      "148512/148512 [==============================] - 86s 581us/sample - loss: 1.4740 - accuracy: 0.5241 - val_loss: 1.5393 - val_accuracy: 0.5164\n",
      "Epoch 3/100\n",
      "148512/148512 [==============================] - 87s 584us/sample - loss: 1.2411 - accuracy: 0.6132 - val_loss: 1.4550 - val_accuracy: 0.5857\n",
      "Epoch 4/100\n",
      "148512/148512 [==============================] - 87s 586us/sample - loss: 1.0914 - accuracy: 0.6672 - val_loss: 1.5193 - val_accuracy: 0.5784\n",
      "Epoch 5/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.9891 - accuracy: 0.7010 - val_loss: 1.3516 - val_accuracy: 0.6318\n",
      "Epoch 6/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.9073 - accuracy: 0.7282 - val_loss: 1.5120 - val_accuracy: 0.5857\n",
      "Epoch 7/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.8492 - accuracy: 0.7464 - val_loss: 1.4807 - val_accuracy: 0.6367\n",
      "Epoch 8/100\n",
      "148512/148512 [==============================] - 87s 586us/sample - loss: 0.8021 - accuracy: 0.7611 - val_loss: 1.6266 - val_accuracy: 0.5796\n",
      "Epoch 9/100\n",
      "148512/148512 [==============================] - 87s 586us/sample - loss: 0.7663 - accuracy: 0.7725 - val_loss: 1.4360 - val_accuracy: 0.6355\n",
      "Epoch 10/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.7213 - accuracy: 0.7850 - val_loss: 1.8624 - val_accuracy: 0.6185\n",
      "Epoch 11/100\n",
      "148512/148512 [==============================] - 87s 584us/sample - loss: 0.6909 - accuracy: 0.7942 - val_loss: 1.5323 - val_accuracy: 0.6355\n",
      "Epoch 12/100\n",
      "148512/148512 [==============================] - 87s 588us/sample - loss: 0.6683 - accuracy: 0.7999 - val_loss: 1.8941 - val_accuracy: 0.5541\n",
      "Epoch 13/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.6426 - accuracy: 0.8073 - val_loss: 1.5787 - val_accuracy: 0.6087\n",
      "Epoch 14/100\n",
      "148512/148512 [==============================] - 87s 585us/sample - loss: 0.6183 - accuracy: 0.8135 - val_loss: 1.7071 - val_accuracy: 0.6549\n",
      "Epoch 15/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.6006 - accuracy: 0.8193 - val_loss: 2.2193 - val_accuracy: 0.5273\n",
      "Epoch 16/100\n",
      "148512/148512 [==============================] - 87s 586us/sample - loss: 0.5841 - accuracy: 0.8255 - val_loss: 1.7476 - val_accuracy: 0.5687\n",
      "Epoch 17/100\n",
      "148512/148512 [==============================] - 87s 586us/sample - loss: 0.5727 - accuracy: 0.8279 - val_loss: 1.8714 - val_accuracy: 0.5869\n",
      "Epoch 18/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.5545 - accuracy: 0.8333 - val_loss: 1.6949 - val_accuracy: 0.6245\n",
      "Epoch 19/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.5419 - accuracy: 0.8363 - val_loss: 1.7884 - val_accuracy: 0.6768\n",
      "Epoch 20/100\n",
      "148512/148512 [==============================] - 87s 587us/sample - loss: 0.5250 - accuracy: 0.8413 - val_loss: 1.9363 - val_accuracy: 0.6525\n",
      "Evaluating model...\n",
      "ROC: 0.959\n",
      "\n",
      "Accuracy = 0.70\n",
      "F-Score: 0.7\n",
      "\n",
      "Average R.O.C: 0.9426131557060643\n",
      "Average Accuracy: 0.6669977366924286\n"
     ]
    }
   ],
   "source": [
    "acc = np.zeros(10)\n",
    "roc = np.zeros(10)\n",
    "\n",
    "CM = 0\n",
    "\n",
    "for f in range(1,10+1):\n",
    "\n",
    "    # load data\n",
    "    train_x, test_x, val_x, train_y, test_y, val_y = load_folds(load_dir,augmented_load_dir, f)\n",
    "    #train_gen, test_x, val_x, test_y, val_y = load_folds_pescador(load_dir,augmented_load_dir, f)\n",
    "\n",
    "    \n",
    "    print(\"Building model...\")\n",
    "    model = build_model(channels = 1)\n",
    "\n",
    "    # now fit the model to the training data, evaluating loss against the validation data\n",
    "    print(\"Training model...\")\n",
    "    model.fit(train_x, train_y, validation_data=(test_x, test_y), \n",
    "              callbacks=[EarlyStopping(restore_best_weights=True, patience=15)],\n",
    "              batch_size=100, epochs=100)\n",
    "    \n",
    "    # now evaluate the trained model against the unseen test data\n",
    "    print(\"Evaluating model...\")\n",
    "    roc[f-1], acc[f-1], cm = evaluate(model, val_x, val_y)\n",
    "    CM += cm\n",
    "\n",
    "    \n",
    "print ('\\nAverage R.O.C:', np.mean(roc))\n",
    "print ('Average Accuracy:', np.mean(acc))\n",
    "\n",
    "# using all folds: best ROC = 0.91, f-score = 0.592 (50 epochs)\n",
    "# using 2 folds: average ROC = 0.792, average f-score = 0.335\n",
    "\n",
    "# if you want to save the model, uncomment this...\n",
    "#filepath = \"models/salamon-cnn-model.h5\"\n",
    "#model.save(filepath)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "^C\n"
     ]
    }
   ],
   "source": [
    "!conda install matplotlib seaborn pandas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting matplotlib\n",
      "  Downloading matplotlib-3.2.1-cp37-cp37m-win_amd64.whl (9.2 MB)\n",
      "Collecting seaborn\n",
      "  Downloading seaborn-0.10.1-py3-none-any.whl (215 kB)\n",
      "Collecting kiwisolver>=1.0.1\n",
      "  Downloading kiwisolver-1.2.0-cp37-none-win_amd64.whl (57 kB)\n",
      "Collecting cycler>=0.10\n",
      "  Downloading cycler-0.10.0-py2.py3-none-any.whl (6.5 kB)\n",
      "Requirement already satisfied: python-dateutil>=2.1 in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from matplotlib) (2.8.1)\n",
      "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from matplotlib) (2.4.7)\n",
      "Requirement already satisfied: numpy>=1.11 in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from matplotlib) (1.18.5)\n",
      "Requirement already satisfied: pandas>=0.22.0 in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from seaborn) (1.0.4)\n",
      "Requirement already satisfied: scipy>=1.0.1 in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from seaborn) (1.4.1)\n",
      "Requirement already satisfied: six in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from cycler>=0.10->matplotlib) (1.15.0)\n",
      "Requirement already satisfied: pytz>=2017.2 in c:\\users\\monic\\miniconda3\\envs\\deepsound\\lib\\site-packages (from pandas>=0.22.0->seaborn) (2020.1)\n",
      "Installing collected packages: kiwisolver, cycler, matplotlib, seaborn\n",
      "Successfully installed cycler-0.10.0 kiwisolver-1.2.0 matplotlib-3.2.1 seaborn-0.10.1\n"
     ]
    }
   ],
   "source": [
    "!pip install matplotlib seaborn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9TlA2tsI5CFF"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'whiskers': [<matplotlib.lines.Line2D at 0x1b18c804c48>,\n",
       "  <matplotlib.lines.Line2D at 0x1b18c804d08>],\n",
       " 'caps': [<matplotlib.lines.Line2D at 0x1b18c80a1c8>,\n",
       "  <matplotlib.lines.Line2D at 0x1b18c80eb48>],\n",
       " 'boxes': [<matplotlib.lines.Line2D at 0x1b18c804a88>],\n",
       " 'medians': [<matplotlib.lines.Line2D at 0x1b18c80e548>],\n",
       " 'fliers': [<matplotlib.lines.Line2D at 0x1b18dcd1888>],\n",
       " 'means': []}"
      ]
     },
     "execution_count": 11,
     "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": [
    "plt.boxplot(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "UgKVz2_a5Cnn"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1b18db05c08>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "classes = ['air_conditioner',\n",
    "'car_horn',\n",
    "'children_playing',\n",
    "'dog_bark',\n",
    "'drilling',\n",
    "'engine_idling',\n",
    "'gun_shot',\n",
    "'jackhammer',\n",
    "'siren',\n",
    "'street_music']\n",
    "df_cm = pd.DataFrame(CM, index = classes,\n",
    "                  columns = classes)\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(df_cm, annot=True,  fmt='d')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.save(\"acc_augmented.npy\", acc)\n",
    "np.save(\"cm_agumented.npy\", CM)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "anaconda-cloud": {},
  "colab": {
   "collapsed_sections": [],
   "name": "us8k-cnn-salamon.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}