Fusion_Training_Augmented_1000ms.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fushion of IMU and Audio features\n",
    "\n",
    "## Training on augmented datasets: 1000ms augmentation case.\n",
    "- Training is done by including: 1-shift (50ms), 2-shifts (100ms), 10-shifts (500ms), 20-shifts (1000ms)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## IMU and Audio data is concatenated at an intermediate layer and then prediction is made"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:523: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:524: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:527: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:532: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import scipy.io\n",
    "\n",
    "from keras import optimizers\n",
    "from keras.optimizers import SGD\n",
    "from keras.optimizers import Adam\n",
    "from keras.metrics import categorical_crossentropy\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.convolutional import *\n",
    "from keras.callbacks import Callback\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "\n",
    "\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D, LSTM, Dense, Dropout, Flatten, Activation\n",
    "from keras.layers.core import Permute, Reshape\n",
    "from keras import backend as K\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "from PIL import Image as img_PIL\n",
    "\n",
    "\n",
    "# random seed.\n",
    "rand_seed = 2\n",
    "\n",
    "from numpy.random import seed\n",
    "seed(rand_seed)\n",
    "from tensorflow import set_random_seed\n",
    "set_random_seed(rand_seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_hot_encoding(y_data):\n",
    "    Mapping=dict()\n",
    "    sub_dirs=['downstair','upstair','run','jump','walk','handwashing','exercise']\n",
    "\n",
    "    categories=7\n",
    "\n",
    "    count=0\n",
    "    for i in sub_dirs:\n",
    "        Mapping[i]=count\n",
    "        count=count+1\n",
    "\n",
    "    y_features2=[]\n",
    "    for i in range(len(y_data)):\n",
    "        Type=y_data[i]\n",
    "        lab=Mapping[Type]\n",
    "        y_features2.append(lab)\n",
    "\n",
    "    y_features=np.array(y_features2)\n",
    "    y_features=y_features.reshape(y_features.shape[0],1)\n",
    "    from keras.utils import to_categorical\n",
    "    y_features = to_categorical(y_features)\n",
    "\n",
    "    return y_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "path='data/'\n",
    "\n",
    "\n",
    "def get_train_data(path=path):\n",
    "    Train_data=np.load(path+'train_5000.npz')\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "\n",
    "    Labels = one_hot_encoding(Labels)\n",
    "    Features_imu = Features_imu.reshape(Features_imu.shape[0],1, Features_imu.shape[1], Features_imu.shape[2]) \n",
    "\n",
    "    return Features_imu,Labels,Features_sound\n",
    "\n",
    "def get_valid_data(path=path):\n",
    "    Train_data=np.load(path+'valid_1000.npz')\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "\n",
    "    Labels = one_hot_encoding(Labels)\n",
    "    Features_imu = Features_imu.reshape(Features_imu.shape[0],1, Features_imu.shape[1], Features_imu.shape[2]) \n",
    "\n",
    "    return Features_imu,Labels,Features_sound\n",
    "\n",
    "def get_test_data(path=path):\n",
    "    Train_data=np.load(path+'test_1377.npz')\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "\n",
    "    Labels = one_hot_encoding(Labels)\n",
    "    Features_imu = Features_imu.reshape(Features_imu.shape[0],1, Features_imu.shape[1], Features_imu.shape[2]) \n",
    "\n",
    "    return Features_imu,Labels,Features_sound\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5000, 1, 40, 12) (5000, 7) (5000, 193)\n",
      "Train Classes distribution:  [760. 994. 546. 780. 532. 732. 656.]\n",
      "(1000, 1, 40, 12) (1000, 7) (1000, 193)\n",
      "Valid Classes distribution:  [150. 188. 136. 141.  98. 157. 130.]\n",
      "(1377, 1, 40, 12) (1377, 7) (1377, 193)\n",
      "Test Classes distribution:  [219. 268. 124. 146. 143. 234. 243.]\n"
     ]
    }
   ],
   "source": [
    "Features_imu1,Labels1,Features_sound1 = get_train_data()\n",
    "print(Features_imu1.shape, Labels1.shape, Features_sound1.shape)\n",
    "print('Train Classes distribution: ',np.sum(Labels1, axis =0))\n",
    "\n",
    "Features_imu2,Labels2,Features_sound2 = get_valid_data()\n",
    "print(Features_imu2.shape, Labels2.shape, Features_sound2.shape)\n",
    "print('Valid Classes distribution: ',np.sum(Labels2, axis =0))\n",
    "\n",
    "Features_imu3,Labels3,Features_sound3 = get_test_data()\n",
    "print(Features_imu3.shape, Labels3.shape, Features_sound3.shape)\n",
    "print('Test Classes distribution: ',np.sum(Labels3, axis =0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_data2(path,file):\n",
    "    Train_data=np.load(path+file)\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "    return Features_imu,Labels,Features_sound\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4999, 1, 40, 12) (4999, 7) (4999, 193)\n",
      "Train Classes distribution:  [760. 994. 546. 780. 532. 732. 655.]\n"
     ]
    }
   ],
   "source": [
    "Features_imu_1,Labels_1,Features_sound_1 = get_train_data2(path,'train_data_1_shift.npz')\n",
    "print(Features_imu_1.shape, Labels_1.shape, Features_sound_1.shape)\n",
    "print('Train Classes distribution: ',np.sum(Labels_1, axis =0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4999, 1, 40, 12) (4999, 7) (4999, 193)\n",
      "Train Classes distribution:  [760. 994. 546. 780. 532. 732. 655.]\n"
     ]
    }
   ],
   "source": [
    "Features_imu_2,Labels_2,Features_sound_2 = get_train_data2(path,'train_data_2_shift.npz')\n",
    "print(Features_imu_2.shape, Labels_2.shape, Features_sound_2.shape)\n",
    "print('Train Classes distribution: ',np.sum(Labels_2, axis =0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4999, 1, 40, 12) (4999, 7) (4999, 193)\n",
      "Train Classes distribution:  [760. 994. 546. 780. 532. 732. 655.]\n"
     ]
    }
   ],
   "source": [
    "Features_imu_10,Labels_10,Features_sound_10 = get_train_data2(path,'train_data_10_shift.npz')\n",
    "print(Features_imu_10.shape, Labels_10.shape, Features_sound_10.shape)\n",
    "print('Train Classes distribution: ',np.sum(Labels_10, axis =0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4999, 1, 40, 12) (4999, 7) (4999, 193)\n",
      "Train Classes distribution:  [760. 994. 546. 780. 532. 732. 655.]\n"
     ]
    }
   ],
   "source": [
    "Features_imu_20,Labels_20,Features_sound_20 = get_train_data2(path,'train_data_20_shift.npz')\n",
    "print(Features_imu_20.shape, Labels_20.shape, Features_sound_20.shape)\n",
    "print('Train Classes distribution: ',np.sum(Labels_20, axis =0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Make augmented Training Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24996, 1, 40, 12)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Features_imu = np.concatenate((Features_imu1,Features_imu_1,Features_imu_2,Features_imu_10,Features_imu_20),axis=0)\n",
    "Features_imu.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24996, 7)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Labels = np.concatenate((Labels1,Labels_1,Labels_2,Labels_10,Labels_20),axis=0)\n",
    "Labels.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(24996, 193)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Features_sound = np.concatenate((Features_sound1,Features_sound_1,Features_sound_2,Features_sound_10,Features_sound_20),axis=0)\n",
    "Features_sound.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_path = 'data/IMU_Audio_Fusion_1000ms'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Combined Model: Audio + IMU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "rand_seed = 2\n",
    "\n",
    "from numpy.random import seed\n",
    "seed(rand_seed)\n",
    "from tensorflow import set_random_seed\n",
    "set_random_seed(rand_seed)\n",
    "\n",
    "input_audio = keras.layers.Input(shape=(193,))\n",
    "A1 = keras.layers.Dense(256, activation='relu')(input_audio)\n",
    "A2 = keras.layers.Dense(256, activation='relu')(A1)\n",
    "D1 = keras.layers.Dropout(0.3)(A2)\n",
    "A5 = keras.layers.Dense(128, activation='relu')(D1)\n",
    "\n",
    "\n",
    "\n",
    "num_feat_map=32\n",
    "_, dim, win_len, _, = Features_imu.shape\n",
    "input_imu = keras.layers.Input(shape=(dim, win_len, 12))\n",
    "\n",
    "I1 = keras.layers.Conv2D(num_feat_map, kernel_size=(1, 5),\n",
    "                 activation='relu',\n",
    "                 padding='same')(input_imu)\n",
    "I2=keras.layers.MaxPooling2D(pool_size=(1, 2))(I1)\n",
    "\n",
    "I3 = keras.layers.Conv2D(num_feat_map, kernel_size=(1, 5),\n",
    "                 activation='relu',\n",
    "                 padding='same')(I2)\n",
    "I4=keras.layers.MaxPooling2D(pool_size=(1, 2))(I3)\n",
    "I5=keras.layers.Flatten()(I4)\n",
    "\n",
    "I6= keras.layers.Dense(256, activation='relu')(I5)\n",
    "D2 = keras.layers.Dropout(0.3)(I6)\n",
    "\n",
    "I7= keras.layers.Dense(128, activation='relu')(D2)\n",
    "\n",
    "\n",
    "# equivalent to added = keras.layers.add([x1, x2])\n",
    "C1 =keras.layers.concatenate([A5, I7], axis=-1)\n",
    "\n",
    "\n",
    "C3= keras.layers.Dense(64, activation='relu')(C1)\n",
    "\n",
    "out = keras.layers.Dense(7, activation='softmax')(C3)\n",
    "\n",
    "model = keras.models.Model(inputs=[input_audio, input_imu], outputs=out)\n",
    "#model = keras.models.Model(inputs=input_audio, outputs=out)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "\n",
    "from keras.callbacks import ReduceLROnPlateau\n",
    "from keras.callbacks import Callback\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=model_path, verbose=1,monitor='val_acc' ,save_best_only=True, save_weights_only=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 24996 samples, validate on 1000 samples\n",
      "Epoch 1/100\n",
      "24996/24996 [==============================] - 40s 2ms/step - loss: 3.4956 - acc: 0.6727 - val_loss: 0.2626 - val_acc: 0.9260\n",
      "\n",
      "Epoch 00001: val_acc improved from -inf to 0.92600, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 2/100\n",
      "24996/24996 [==============================] - 36s 1ms/step - loss: 0.1964 - acc: 0.9322 - val_loss: 0.1321 - val_acc: 0.9640\n",
      "\n",
      "Epoch 00002: val_acc improved from 0.92600 to 0.96400, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 3/100\n",
      "24996/24996 [==============================] - 36s 1ms/step - loss: 0.1112 - acc: 0.9582 - val_loss: 0.0998 - val_acc: 0.9710\n",
      "\n",
      "Epoch 00003: val_acc improved from 0.96400 to 0.97100, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 4/100\n",
      "24996/24996 [==============================] - 25s 1ms/step - loss: 0.0720 - acc: 0.9728 - val_loss: 0.1093 - val_acc: 0.9680\n",
      "\n",
      "Epoch 00004: val_acc did not improve from 0.97100\n",
      "Epoch 5/100\n",
      "24996/24996 [==============================] - 21s 836us/step - loss: 0.0649 - acc: 0.9769 - val_loss: 0.1027 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00005: val_acc improved from 0.97100 to 0.97400, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 6/100\n",
      "24996/24996 [==============================] - 22s 898us/step - loss: 0.0461 - acc: 0.9841 - val_loss: 0.0692 - val_acc: 0.9770\n",
      "\n",
      "Epoch 00006: val_acc improved from 0.97400 to 0.97700, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 7/100\n",
      "24996/24996 [==============================] - 21s 857us/step - loss: 0.0405 - acc: 0.9848 - val_loss: 0.1022 - val_acc: 0.9700\n",
      "\n",
      "Epoch 00007: val_acc did not improve from 0.97700\n",
      "Epoch 8/100\n",
      "24996/24996 [==============================] - 22s 874us/step - loss: 0.0411 - acc: 0.9848 - val_loss: 0.1346 - val_acc: 0.9630\n",
      "\n",
      "Epoch 00008: val_acc did not improve from 0.97700\n",
      "Epoch 9/100\n",
      "24996/24996 [==============================] - 21s 847us/step - loss: 0.0428 - acc: 0.9848 - val_loss: 0.0938 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00009: val_acc did not improve from 0.97700\n",
      "Epoch 10/100\n",
      "24996/24996 [==============================] - 21s 831us/step - loss: 0.0295 - acc: 0.9893 - val_loss: 0.0740 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00010: val_acc improved from 0.97700 to 0.98000, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 11/100\n",
      "24996/24996 [==============================] - 20s 814us/step - loss: 0.0267 - acc: 0.9906 - val_loss: 0.0727 - val_acc: 0.9840\n",
      "\n",
      "Epoch 00011: val_acc improved from 0.98000 to 0.98400, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 12/100\n",
      "24996/24996 [==============================] - 11s 449us/step - loss: 0.0194 - acc: 0.9933 - val_loss: 0.1129 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00012: val_acc did not improve from 0.98400\n",
      "Epoch 13/100\n",
      "24996/24996 [==============================] - 12s 467us/step - loss: 0.0180 - acc: 0.9937 - val_loss: 0.1512 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00013: val_acc did not improve from 0.98400\n",
      "Epoch 14/100\n",
      "24996/24996 [==============================] - 12s 463us/step - loss: 0.0407 - acc: 0.9874 - val_loss: 0.1212 - val_acc: 0.9720\n",
      "\n",
      "Epoch 00014: val_acc did not improve from 0.98400\n",
      "Epoch 15/100\n",
      "24996/24996 [==============================] - 12s 484us/step - loss: 0.0241 - acc: 0.9920 - val_loss: 0.0845 - val_acc: 0.9810\n",
      "\n",
      "Epoch 00015: val_acc did not improve from 0.98400\n",
      "Epoch 16/100\n",
      "24996/24996 [==============================] - 12s 475us/step - loss: 0.0199 - acc: 0.9937 - val_loss: 0.0841 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00016: val_acc did not improve from 0.98400\n",
      "Epoch 17/100\n",
      "24996/24996 [==============================] - 12s 460us/step - loss: 0.0252 - acc: 0.9912 - val_loss: 0.0983 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00017: val_acc improved from 0.98400 to 0.98500, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 18/100\n",
      "24996/24996 [==============================] - 11s 450us/step - loss: 0.0227 - acc: 0.9936 - val_loss: 0.1850 - val_acc: 0.9670\n",
      "\n",
      "Epoch 00018: val_acc did not improve from 0.98500\n",
      "Epoch 19/100\n",
      "24996/24996 [==============================] - 12s 462us/step - loss: 0.0420 - acc: 0.9885 - val_loss: 0.1569 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00019: val_acc did not improve from 0.98500\n",
      "Epoch 20/100\n",
      "24996/24996 [==============================] - 12s 464us/step - loss: 0.0230 - acc: 0.9936 - val_loss: 0.1098 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00020: val_acc did not improve from 0.98500\n",
      "Epoch 21/100\n",
      "24996/24996 [==============================] - 11s 457us/step - loss: 0.0243 - acc: 0.9936 - val_loss: 0.1011 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00021: val_acc did not improve from 0.98500\n",
      "Epoch 22/100\n",
      "24996/24996 [==============================] - 11s 457us/step - loss: 0.0349 - acc: 0.9910 - val_loss: 0.0852 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00022: val_acc did not improve from 0.98500\n",
      "Epoch 23/100\n",
      "24996/24996 [==============================] - 11s 458us/step - loss: 0.0161 - acc: 0.9953 - val_loss: 0.0819 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00023: val_acc improved from 0.98500 to 0.98700, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 24/100\n",
      "24996/24996 [==============================] - 11s 442us/step - loss: 0.0047 - acc: 0.9986 - val_loss: 0.0682 - val_acc: 0.9910\n",
      "\n",
      "Epoch 00024: val_acc improved from 0.98700 to 0.99100, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 25/100\n",
      "24996/24996 [==============================] - 11s 446us/step - loss: 0.0193 - acc: 0.9954 - val_loss: 0.1361 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00025: val_acc did not improve from 0.99100\n",
      "Epoch 26/100\n",
      "24996/24996 [==============================] - 11s 448us/step - loss: 0.0394 - acc: 0.9902 - val_loss: 0.0504 - val_acc: 0.9920\n",
      "\n",
      "Epoch 00026: val_acc improved from 0.99100 to 0.99200, saving model to data/IMU_Audio_Fusion_1000ms\n",
      "Epoch 27/100\n",
      "24996/24996 [==============================] - 11s 457us/step - loss: 0.0221 - acc: 0.9943 - val_loss: 0.1072 - val_acc: 0.9810\n",
      "\n",
      "Epoch 00027: val_acc did not improve from 0.99200\n",
      "Epoch 28/100\n",
      "24996/24996 [==============================] - 11s 458us/step - loss: 0.0153 - acc: 0.9956 - val_loss: 0.0936 - val_acc: 0.9810\n",
      "\n",
      "Epoch 00028: val_acc did not improve from 0.99200\n",
      "Epoch 29/100\n",
      "24996/24996 [==============================] - 11s 454us/step - loss: 0.0074 - acc: 0.9976 - val_loss: 0.0727 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00029: val_acc did not improve from 0.99200\n",
      "Epoch 30/100\n",
      "24996/24996 [==============================] - 11s 445us/step - loss: 0.0049 - acc: 0.9987 - val_loss: 0.0844 - val_acc: 0.9810\n",
      "\n",
      "Epoch 00030: val_acc did not improve from 0.99200\n",
      "Epoch 31/100\n",
      "24996/24996 [==============================] - 12s 488us/step - loss: 0.0179 - acc: 0.9954 - val_loss: 0.1220 - val_acc: 0.9760\n",
      "\n",
      "Epoch 00031: val_acc did not improve from 0.99200\n",
      "Epoch 32/100\n",
      "24996/24996 [==============================] - 13s 514us/step - loss: 0.0167 - acc: 0.9955 - val_loss: 0.0839 - val_acc: 0.9860\n",
      "\n",
      "Epoch 00032: val_acc did not improve from 0.99200\n",
      "Epoch 33/100\n",
      "24996/24996 [==============================] - 13s 506us/step - loss: 0.0127 - acc: 0.9967 - val_loss: 0.0800 - val_acc: 0.9840\n",
      "\n",
      "Epoch 00033: val_acc did not improve from 0.99200\n",
      "Epoch 34/100\n",
      "24996/24996 [==============================] - 12s 477us/step - loss: 0.0084 - acc: 0.9974 - val_loss: 0.0536 - val_acc: 0.9880\n",
      "\n",
      "Epoch 00034: val_acc did not improve from 0.99200\n",
      "Epoch 35/100\n",
      "24996/24996 [==============================] - 12s 477us/step - loss: 0.0423 - acc: 0.9913 - val_loss: 0.1814 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00035: val_acc did not improve from 0.99200\n",
      "Epoch 36/100\n",
      "24996/24996 [==============================] - 12s 474us/step - loss: 0.0258 - acc: 0.9938 - val_loss: 0.1211 - val_acc: 0.9760\n",
      "\n",
      "Epoch 00036: val_acc did not improve from 0.99200\n",
      "Epoch 37/100\n",
      "24996/24996 [==============================] - 12s 483us/step - loss: 0.0218 - acc: 0.9941 - val_loss: 0.0741 - val_acc: 0.9890\n",
      "\n",
      "Epoch 00037: val_acc did not improve from 0.99200\n",
      "Epoch 38/100\n",
      "24996/24996 [==============================] - 12s 481us/step - loss: 0.0067 - acc: 0.9985 - val_loss: 0.1404 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00038: val_acc did not improve from 0.99200\n",
      "Epoch 39/100\n",
      "24996/24996 [==============================] - 12s 472us/step - loss: 0.0167 - acc: 0.9958 - val_loss: 0.0692 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00039: val_acc did not improve from 0.99200\n",
      "Epoch 40/100\n",
      "24996/24996 [==============================] - 13s 502us/step - loss: 0.0247 - acc: 0.9943 - val_loss: 0.2019 - val_acc: 0.9680\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00040: val_acc did not improve from 0.99200\n",
      "Epoch 41/100\n",
      "24996/24996 [==============================] - 11s 440us/step - loss: 0.0201 - acc: 0.9952 - val_loss: 0.1135 - val_acc: 0.9840\n",
      "\n",
      "Epoch 00041: val_acc did not improve from 0.99200\n",
      "Epoch 42/100\n",
      "24996/24996 [==============================] - 11s 437us/step - loss: 0.0090 - acc: 0.9971 - val_loss: 0.0773 - val_acc: 0.9890\n",
      "\n",
      "Epoch 00042: val_acc did not improve from 0.99200\n",
      "Epoch 43/100\n",
      "24996/24996 [==============================] - 11s 450us/step - loss: 0.0072 - acc: 0.9983 - val_loss: 0.1133 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00043: val_acc did not improve from 0.99200\n",
      "Epoch 44/100\n",
      "24996/24996 [==============================] - 12s 476us/step - loss: 0.0177 - acc: 0.9963 - val_loss: 0.0959 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00044: val_acc did not improve from 0.99200\n",
      "Epoch 45/100\n",
      "24996/24996 [==============================] - 12s 465us/step - loss: 0.0037 - acc: 0.9989 - val_loss: 0.0819 - val_acc: 0.9880\n",
      "\n",
      "Epoch 00045: val_acc did not improve from 0.99200\n",
      "Epoch 46/100\n",
      "24996/24996 [==============================] - 12s 482us/step - loss: 0.0171 - acc: 0.9959 - val_loss: 0.0896 - val_acc: 0.9840\n",
      "\n",
      "Epoch 00046: val_acc did not improve from 0.99200\n",
      "Epoch 47/100\n",
      "24996/24996 [==============================] - 12s 468us/step - loss: 0.0155 - acc: 0.9970 - val_loss: 0.1244 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00047: val_acc did not improve from 0.99200\n",
      "Epoch 48/100\n",
      "24996/24996 [==============================] - 12s 461us/step - loss: 0.0091 - acc: 0.9978 - val_loss: 0.0573 - val_acc: 0.9890\n",
      "\n",
      "Epoch 00048: val_acc did not improve from 0.99200\n",
      "Epoch 49/100\n",
      "24996/24996 [==============================] - 12s 466us/step - loss: 0.0197 - acc: 0.9962 - val_loss: 0.1050 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00049: val_acc did not improve from 0.99200\n",
      "Epoch 50/100\n",
      "24996/24996 [==============================] - 15s 582us/step - loss: 0.0284 - acc: 0.9941 - val_loss: 0.0906 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00050: val_acc did not improve from 0.99200\n",
      "Epoch 51/100\n",
      "24996/24996 [==============================] - 15s 590us/step - loss: 0.0215 - acc: 0.9961 - val_loss: 0.0732 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00051: val_acc did not improve from 0.99200\n",
      "Epoch 52/100\n",
      "24996/24996 [==============================] - 15s 615us/step - loss: 0.0093 - acc: 0.9978 - val_loss: 0.0880 - val_acc: 0.9860\n",
      "\n",
      "Epoch 00052: val_acc did not improve from 0.99200\n",
      "Epoch 53/100\n",
      "24996/24996 [==============================] - 12s 478us/step - loss: 0.0210 - acc: 0.9962 - val_loss: 0.0894 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00053: val_acc did not improve from 0.99200\n",
      "Epoch 54/100\n",
      "24996/24996 [==============================] - 12s 487us/step - loss: 0.0159 - acc: 0.9965 - val_loss: 0.0710 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00054: val_acc did not improve from 0.99200\n",
      "Epoch 55/100\n",
      "24996/24996 [==============================] - 12s 469us/step - loss: 0.0061 - acc: 0.9983 - val_loss: 0.1023 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00055: val_acc did not improve from 0.99200\n",
      "Epoch 56/100\n",
      "24996/24996 [==============================] - 12s 468us/step - loss: 0.0058 - acc: 0.9988 - val_loss: 0.1614 - val_acc: 0.9750\n",
      "\n",
      "Epoch 00056: val_acc did not improve from 0.99200\n",
      "Epoch 57/100\n",
      "24996/24996 [==============================] - 12s 491us/step - loss: 0.0205 - acc: 0.9960 - val_loss: 0.1031 - val_acc: 0.9840\n",
      "\n",
      "Epoch 00057: val_acc did not improve from 0.99200\n",
      "Epoch 58/100\n",
      "24996/24996 [==============================] - 14s 545us/step - loss: 0.0067 - acc: 0.9983 - val_loss: 0.1066 - val_acc: 0.9880\n",
      "\n",
      "Epoch 00058: val_acc did not improve from 0.99200\n",
      "Epoch 59/100\n",
      "24996/24996 [==============================] - 11s 452us/step - loss: 0.0035 - acc: 0.9989 - val_loss: 0.1645 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00059: val_acc did not improve from 0.99200\n",
      "Epoch 60/100\n",
      "24996/24996 [==============================] - 12s 466us/step - loss: 0.0115 - acc: 0.9975 - val_loss: 0.1155 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00060: val_acc did not improve from 0.99200\n",
      "Epoch 61/100\n",
      "24996/24996 [==============================] - 15s 607us/step - loss: 0.0211 - acc: 0.9957 - val_loss: 0.1104 - val_acc: 0.9860\n",
      "\n",
      "Epoch 00061: val_acc did not improve from 0.99200\n",
      "Epoch 62/100\n",
      "24996/24996 [==============================] - 22s 864us/step - loss: 0.0148 - acc: 0.9966 - val_loss: 0.0932 - val_acc: 0.9860\n",
      "\n",
      "Epoch 00062: val_acc did not improve from 0.99200\n",
      "Epoch 63/100\n",
      "24996/24996 [==============================] - 21s 859us/step - loss: 0.0189 - acc: 0.9955 - val_loss: 0.0733 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00063: val_acc did not improve from 0.99200\n",
      "Epoch 64/100\n",
      "24996/24996 [==============================] - 21s 846us/step - loss: 0.0053 - acc: 0.9988 - val_loss: 0.0644 - val_acc: 0.9910\n",
      "\n",
      "Epoch 00064: val_acc did not improve from 0.99200\n",
      "Epoch 65/100\n",
      "24996/24996 [==============================] - 21s 860us/step - loss: 0.0045 - acc: 0.9988 - val_loss: 0.0905 - val_acc: 0.9900\n",
      "\n",
      "Epoch 00065: val_acc did not improve from 0.99200\n",
      "Epoch 66/100\n",
      "24996/24996 [==============================] - 22s 873us/step - loss: 0.0028 - acc: 0.9994 - val_loss: 0.0905 - val_acc: 0.9910\n",
      "\n",
      "Epoch 00066: val_acc did not improve from 0.99200\n",
      "Epoch 67/100\n",
      "24996/24996 [==============================] - 21s 838us/step - loss: 0.0056 - acc: 0.9992 - val_loss: 0.1651 - val_acc: 0.9810\n",
      "\n",
      "Epoch 00067: val_acc did not improve from 0.99200\n",
      "Epoch 68/100\n",
      "24996/24996 [==============================] - 19s 761us/step - loss: 0.0160 - acc: 0.9972 - val_loss: 0.0872 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00068: val_acc did not improve from 0.99200\n",
      "Epoch 69/100\n",
      "24996/24996 [==============================] - 22s 864us/step - loss: 0.0180 - acc: 0.9966 - val_loss: 0.2650 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00069: val_acc did not improve from 0.99200\n",
      "Epoch 70/100\n",
      "24996/24996 [==============================] - 21s 859us/step - loss: 0.0448 - acc: 0.9927 - val_loss: 0.2031 - val_acc: 0.9720\n",
      "\n",
      "Epoch 00070: val_acc did not improve from 0.99200\n",
      "Epoch 71/100\n",
      "24996/24996 [==============================] - 21s 824us/step - loss: 0.0241 - acc: 0.9959 - val_loss: 0.1029 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00071: val_acc did not improve from 0.99200\n",
      "Epoch 72/100\n",
      "24996/24996 [==============================] - 21s 823us/step - loss: 0.0179 - acc: 0.9979 - val_loss: 0.1116 - val_acc: 0.9840\n",
      "\n",
      "Epoch 00072: val_acc did not improve from 0.99200\n",
      "Epoch 73/100\n",
      "24996/24996 [==============================] - 21s 850us/step - loss: 0.0057 - acc: 0.9987 - val_loss: 0.1872 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00073: val_acc did not improve from 0.99200\n",
      "Epoch 74/100\n",
      "24996/24996 [==============================] - 21s 821us/step - loss: 0.0213 - acc: 0.9964 - val_loss: 0.1311 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00074: val_acc did not improve from 0.99200\n",
      "Epoch 75/100\n",
      "24996/24996 [==============================] - 21s 847us/step - loss: 0.0166 - acc: 0.9977 - val_loss: 0.1464 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00075: val_acc did not improve from 0.99200\n",
      "Epoch 76/100\n",
      "24996/24996 [==============================] - 21s 852us/step - loss: 0.0377 - acc: 0.9949 - val_loss: 0.1209 - val_acc: 0.9830\n",
      "\n",
      "Epoch 00076: val_acc did not improve from 0.99200\n",
      "Epoch 77/100\n",
      "24996/24996 [==============================] - 21s 841us/step - loss: 0.0052 - acc: 0.9988 - val_loss: 0.1139 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00077: val_acc did not improve from 0.99200\n",
      "Epoch 78/100\n",
      "24996/24996 [==============================] - 21s 822us/step - loss: 0.0020 - acc: 0.9996 - val_loss: 0.0976 - val_acc: 0.9860\n",
      "\n",
      "Epoch 00078: val_acc did not improve from 0.99200\n",
      "Epoch 79/100\n",
      "24996/24996 [==============================] - 21s 852us/step - loss: 0.0133 - acc: 0.9977 - val_loss: 0.0839 - val_acc: 0.9890\n",
      "\n",
      "Epoch 00079: val_acc did not improve from 0.99200\n",
      "Epoch 80/100\n",
      "24996/24996 [==============================] - 21s 826us/step - loss: 0.0140 - acc: 0.9978 - val_loss: 0.1584 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00080: val_acc did not improve from 0.99200\n",
      "Epoch 81/100\n",
      "24996/24996 [==============================] - 20s 807us/step - loss: 0.0158 - acc: 0.9974 - val_loss: 0.1033 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00081: val_acc did not improve from 0.99200\n",
      "Epoch 82/100\n",
      "24996/24996 [==============================] - 21s 834us/step - loss: 0.0299 - acc: 0.9944 - val_loss: 0.1138 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00082: val_acc did not improve from 0.99200\n",
      "Epoch 83/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24996/24996 [==============================] - 22s 866us/step - loss: 0.0191 - acc: 0.9968 - val_loss: 0.1168 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00083: val_acc did not improve from 0.99200\n",
      "Epoch 84/100\n",
      "24996/24996 [==============================] - 21s 852us/step - loss: 0.0170 - acc: 0.9972 - val_loss: 0.1768 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00084: val_acc did not improve from 0.99200\n",
      "Epoch 85/100\n",
      "24996/24996 [==============================] - 21s 836us/step - loss: 0.0325 - acc: 0.9959 - val_loss: 0.1659 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00085: val_acc did not improve from 0.99200\n",
      "Epoch 86/100\n",
      "24996/24996 [==============================] - 21s 831us/step - loss: 0.0087 - acc: 0.9984 - val_loss: 0.1610 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00086: val_acc did not improve from 0.99200\n",
      "Epoch 87/100\n",
      "24996/24996 [==============================] - 21s 837us/step - loss: 0.0131 - acc: 0.9976 - val_loss: 0.2159 - val_acc: 0.9750\n",
      "\n",
      "Epoch 00087: val_acc did not improve from 0.99200\n",
      "Epoch 88/100\n",
      "24996/24996 [==============================] - 21s 829us/step - loss: 0.0085 - acc: 0.9986 - val_loss: 0.1140 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00088: val_acc did not improve from 0.99200\n",
      "Epoch 89/100\n",
      "24996/24996 [==============================] - 21s 846us/step - loss: 0.0194 - acc: 0.9978 - val_loss: 0.1584 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00089: val_acc did not improve from 0.99200\n",
      "Epoch 90/100\n",
      "24996/24996 [==============================] - 21s 830us/step - loss: 0.0029 - acc: 0.9993 - val_loss: 0.1186 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00090: val_acc did not improve from 0.99200\n",
      "Epoch 91/100\n",
      "24996/24996 [==============================] - 21s 848us/step - loss: 0.0335 - acc: 0.9945 - val_loss: 0.1598 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00091: val_acc did not improve from 0.99200\n",
      "Epoch 92/100\n",
      "24996/24996 [==============================] - 21s 843us/step - loss: 0.0284 - acc: 0.9962 - val_loss: 0.1286 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00092: val_acc did not improve from 0.99200\n",
      "Epoch 93/100\n",
      "24996/24996 [==============================] - 21s 854us/step - loss: 0.0056 - acc: 0.9987 - val_loss: 0.1508 - val_acc: 0.9820\n",
      "\n",
      "Epoch 00093: val_acc did not improve from 0.99200\n",
      "Epoch 94/100\n",
      "24996/24996 [==============================] - 22s 867us/step - loss: 0.0353 - acc: 0.9958 - val_loss: 0.1024 - val_acc: 0.9880\n",
      "\n",
      "Epoch 00094: val_acc did not improve from 0.99200\n",
      "Epoch 95/100\n",
      "24996/24996 [==============================] - 21s 854us/step - loss: 0.0051 - acc: 0.9991 - val_loss: 0.0884 - val_acc: 0.9880\n",
      "\n",
      "Epoch 00095: val_acc did not improve from 0.99200\n",
      "Epoch 96/100\n",
      "24996/24996 [==============================] - 21s 858us/step - loss: 0.0333 - acc: 0.9964 - val_loss: 0.1528 - val_acc: 0.9810\n",
      "\n",
      "Epoch 00096: val_acc did not improve from 0.99200\n",
      "Epoch 97/100\n",
      "24996/24996 [==============================] - 21s 850us/step - loss: 0.0237 - acc: 0.9970 - val_loss: 0.1171 - val_acc: 0.9870\n",
      "\n",
      "Epoch 00097: val_acc did not improve from 0.99200\n",
      "Epoch 98/100\n",
      "24996/24996 [==============================] - 21s 848us/step - loss: 0.0157 - acc: 0.9971 - val_loss: 0.1108 - val_acc: 0.9880\n",
      "\n",
      "Epoch 00098: val_acc did not improve from 0.99200\n",
      "Epoch 99/100\n",
      "24996/24996 [==============================] - 21s 847us/step - loss: 0.0192 - acc: 0.9964 - val_loss: 0.0914 - val_acc: 0.9860\n",
      "\n",
      "Epoch 00099: val_acc did not improve from 0.99200\n",
      "Epoch 100/100\n",
      "24996/24996 [==============================] - 12s 487us/step - loss: 0.0055 - acc: 0.9990 - val_loss: 0.1039 - val_acc: 0.9850\n",
      "\n",
      "Epoch 00100: val_acc did not improve from 0.99200\n",
      "---Time 1699.0373075008392 seconds ---\n"
     ]
    }
   ],
   "source": [
    "batch_size = 128\n",
    "epochs = 100\n",
    "from keras import optimizers\n",
    "from keras.optimizers import Adam\n",
    "\n",
    "adam = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer= adam,\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "import time\n",
    "start_time = time.time()\n",
    "\n",
    "learning_hist = model.fit([Features_sound,Features_imu], Labels,\n",
    "                    verbose=1,\n",
    "                    epochs=epochs,\n",
    "                        shuffle=True,\n",
    "                    batch_size=batch_size,\n",
    "                        callbacks=[checkpointer],\n",
    "                    validation_data=([Features_sound2,Features_imu2], Labels2)\n",
    "                         \n",
    "                         )\n",
    "\n",
    "end_time = time.time()\n",
    "\n",
    "print(\"---Time %s seconds ---\" % (end_time - start_time))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1377/1377 [==============================] - 1s 475us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.17568565193563748, 0.971677559912854]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate([Features_sound3,Features_imu3],Labels3)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "\n",
    "model_fused = load_model(model_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24996/24996 [==============================] - 11s 434us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.006903310225479172, 0.9977596415426468]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Train Accuracy\n",
    "model_fused.evaluate([Features_sound,Features_imu],Labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1000/1000 [==============================] - 0s 414us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.05038352374028546, 0.992]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Valid Accuracy\n",
    "model_fused.evaluate([Features_sound2,Features_imu2],Labels2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1377/1377 [==============================] - 1s 417us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.10806767817294642, 0.9724037763253449]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Test Accuracy\n",
    "model_fused.evaluate([Features_sound3,Features_imu3],Labels3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plotting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "\n",
    "print(learning_hist.history.keys())\n",
    "plt.rcParams.update({'font.size': 18})\n",
    "\n",
    "plt.figure(figsize=(10,5))\n",
    "\n",
    "plt.plot(learning_hist.history['acc'])\n",
    "plt.plot(learning_hist.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'Valid'], loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for loss\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.rcParams.update({'font.size': 18})\n",
    "\n",
    "\n",
    "plt.plot(learning_hist.history['loss'])\n",
    "plt.plot(learning_hist.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'Valid'], loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}