LSTM_v2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas\n",
    "import math\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import LSTM\n",
    "from tensorflow.keras.layers import Dropout\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.metrics import mean_squared_error"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((314, 270), (314, 5))"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_x = np.loadtxt('gesture_input_x_v11.csv', dtype=float)\n",
    "data_y = np.loadtxt('gesture_input_y_v11.csv', dtype=float)\n",
    "\n",
    "data_x.shape, data_y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# fix random seed for reproducibility\n",
    "np.random.seed(7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(251, 63)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# split into train and test sets\n",
    "train_size = int(len(data_x) * 0.8)\n",
    "test_size = len(data_x) - train_size\n",
    "train_size, test_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x, test_x = data_x[0:train_size,:], data_x[train_size:,:]\n",
    "train_y, test_y = data_y[0:train_size,:], data_y[train_size:,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((251, 270), (251, 5), (63, 270), (63, 5))"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_x.shape, train_y.shape, test_x.shape, test_y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# reshape input to be [samples, time steps, features]\n",
    "train_x = np.reshape(train_x, (train_x.shape[0], 1, train_x.shape[1]))\n",
    "test_x = np.reshape(test_x, (test_x.shape[0], 1, test_x.shape[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "WARNING: The TensorFlow contrib module will not be included in TensorFlow 2.0.\n",
      "For more information, please see:\n",
      "  * https://github.com/tensorflow/community/blob/master/rfcs/20180907-contrib-sunset.md\n",
      "  * https://github.com/tensorflow/addons\n",
      "If you depend on functionality not listed there, please file an issue.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "regularizer = tf.contrib.layers.l2_regularizer(scale=0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/darshanakg/.conda/envs/tf_gpu/lib/python3.6/site-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From /home/darshanakg/.conda/envs/tf_gpu/lib/python3.6/site-packages/tensorflow/python/keras/utils/losses_utils.py:170: to_float (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm (LSTM)                  (None, 270)               584280    \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 100)               27100     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 5)                 505       \n",
      "=================================================================\n",
      "Total params: 611,885\n",
      "Trainable params: 611,885\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# create and fit the LSTM network\n",
    "model = Sequential()\n",
    "model.add(LSTM(270, activation=tf.nn.tanh, kernel_regularizer=regularizer, input_shape=(1, 270)))\n",
    "model.add(Dense(100, activation=tf.nn.relu, kernel_regularizer=regularizer))\n",
    "model.add(Dense(5, activation=tf.nn.softmax))\n",
    "\n",
    "model.compile(loss='mean_squared_error', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/darshanakg/.conda/envs/tf_gpu/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Epoch 1/200\n",
      "251/251 [==============================] - 1s 3ms/sample - loss: 2.0046 - acc: 0.3426\n",
      "Epoch 2/200\n",
      "251/251 [==============================] - 0s 405us/sample - loss: 0.7586 - acc: 0.3825\n",
      "Epoch 3/200\n",
      "251/251 [==============================] - 0s 417us/sample - loss: 0.3598 - acc: 0.3865\n",
      "Epoch 4/200\n",
      "251/251 [==============================] - 0s 419us/sample - loss: 0.2329 - acc: 0.3825\n",
      "Epoch 5/200\n",
      "251/251 [==============================] - 0s 417us/sample - loss: 0.1855 - acc: 0.3705\n",
      "Epoch 6/200\n",
      "251/251 [==============================] - 0s 466us/sample - loss: 0.1653 - acc: 0.4263\n",
      "Epoch 7/200\n",
      "251/251 [==============================] - 0s 480us/sample - loss: 0.1544 - acc: 0.4303\n",
      "Epoch 8/200\n",
      "251/251 [==============================] - 0s 501us/sample - loss: 0.1489 - acc: 0.5378\n",
      "Epoch 9/200\n",
      "251/251 [==============================] - 0s 505us/sample - loss: 0.1488 - acc: 0.4661\n",
      "Epoch 10/200\n",
      "251/251 [==============================] - 0s 490us/sample - loss: 0.1405 - acc: 0.5737\n",
      "Epoch 11/200\n",
      "251/251 [==============================] - 0s 478us/sample - loss: 0.1397 - acc: 0.5817\n",
      "Epoch 12/200\n",
      "251/251 [==============================] - 0s 483us/sample - loss: 0.1342 - acc: 0.5976\n",
      "Epoch 13/200\n",
      "251/251 [==============================] - 0s 501us/sample - loss: 0.1322 - acc: 0.5976\n",
      "Epoch 14/200\n",
      "251/251 [==============================] - 0s 493us/sample - loss: 0.1314 - acc: 0.6295\n",
      "Epoch 15/200\n",
      "251/251 [==============================] - 0s 492us/sample - loss: 0.1286 - acc: 0.6375\n",
      "Epoch 16/200\n",
      "251/251 [==============================] - 0s 479us/sample - loss: 0.1344 - acc: 0.6056\n",
      "Epoch 17/200\n",
      "251/251 [==============================] - 0s 491us/sample - loss: 0.1286 - acc: 0.6135\n",
      "Epoch 18/200\n",
      "251/251 [==============================] - 0s 482us/sample - loss: 0.1267 - acc: 0.6614\n",
      "Epoch 19/200\n",
      "251/251 [==============================] - 0s 448us/sample - loss: 0.1281 - acc: 0.6135\n",
      "Epoch 20/200\n",
      "251/251 [==============================] - 0s 499us/sample - loss: 0.1241 - acc: 0.6414\n",
      "Epoch 21/200\n",
      "251/251 [==============================] - 0s 480us/sample - loss: 0.1311 - acc: 0.5936\n",
      "Epoch 22/200\n",
      "251/251 [==============================] - 0s 484us/sample - loss: 0.1234 - acc: 0.6414\n",
      "Epoch 23/200\n",
      "251/251 [==============================] - 0s 472us/sample - loss: 0.1237 - acc: 0.6335\n",
      "Epoch 24/200\n",
      "251/251 [==============================] - 0s 532us/sample - loss: 0.1240 - acc: 0.6574\n",
      "Epoch 25/200\n",
      "251/251 [==============================] - 0s 450us/sample - loss: 0.1186 - acc: 0.6614\n",
      "Epoch 26/200\n",
      "251/251 [==============================] - 0s 421us/sample - loss: 0.1175 - acc: 0.6653\n",
      "Epoch 27/200\n",
      "251/251 [==============================] - 0s 419us/sample - loss: 0.1236 - acc: 0.6335\n",
      "Epoch 28/200\n",
      "251/251 [==============================] - 0s 416us/sample - loss: 0.1237 - acc: 0.6693\n",
      "Epoch 29/200\n",
      "251/251 [==============================] - 0s 458us/sample - loss: 0.1172 - acc: 0.6693\n",
      "Epoch 30/200\n",
      "251/251 [==============================] - 0s 458us/sample - loss: 0.1146 - acc: 0.6693\n",
      "Epoch 31/200\n",
      "251/251 [==============================] - 0s 432us/sample - loss: 0.1223 - acc: 0.6335\n",
      "Epoch 32/200\n",
      "251/251 [==============================] - 0s 414us/sample - loss: 0.1141 - acc: 0.6813\n",
      "Epoch 33/200\n",
      "251/251 [==============================] - 0s 429us/sample - loss: 0.1147 - acc: 0.6653\n",
      "Epoch 34/200\n",
      "251/251 [==============================] - 0s 418us/sample - loss: 0.1100 - acc: 0.7171\n",
      "Epoch 35/200\n",
      "251/251 [==============================] - 0s 499us/sample - loss: 0.1080 - acc: 0.7052\n",
      "Epoch 36/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.1108 - acc: 0.7052\n",
      "Epoch 37/200\n",
      "251/251 [==============================] - 0s 496us/sample - loss: 0.1092 - acc: 0.6813\n",
      "Epoch 38/200\n",
      "251/251 [==============================] - 0s 545us/sample - loss: 0.1086 - acc: 0.7052\n",
      "Epoch 39/200\n",
      "251/251 [==============================] - 0s 426us/sample - loss: 0.1075 - acc: 0.7211\n",
      "Epoch 40/200\n",
      "251/251 [==============================] - 0s 456us/sample - loss: 0.1097 - acc: 0.6773\n",
      "Epoch 41/200\n",
      "251/251 [==============================] - 0s 471us/sample - loss: 0.1148 - acc: 0.6693\n",
      "Epoch 42/200\n",
      "251/251 [==============================] - 0s 496us/sample - loss: 0.1110 - acc: 0.7092\n",
      "Epoch 43/200\n",
      "251/251 [==============================] - 0s 471us/sample - loss: 0.1100 - acc: 0.7131\n",
      "Epoch 44/200\n",
      "251/251 [==============================] - 0s 429us/sample - loss: 0.1123 - acc: 0.6932\n",
      "Epoch 45/200\n",
      "251/251 [==============================] - 0s 438us/sample - loss: 0.1068 - acc: 0.7092\n",
      "Epoch 46/200\n",
      "251/251 [==============================] - 0s 467us/sample - loss: 0.1064 - acc: 0.7251\n",
      "Epoch 47/200\n",
      "251/251 [==============================] - 0s 453us/sample - loss: 0.1032 - acc: 0.6972\n",
      "Epoch 48/200\n",
      "251/251 [==============================] - 0s 505us/sample - loss: 0.1022 - acc: 0.7171\n",
      "Epoch 49/200\n",
      "251/251 [==============================] - 0s 462us/sample - loss: 0.1063 - acc: 0.7131\n",
      "Epoch 50/200\n",
      "251/251 [==============================] - 0s 497us/sample - loss: 0.1000 - acc: 0.7331\n",
      "Epoch 51/200\n",
      "251/251 [==============================] - 0s 475us/sample - loss: 0.1053 - acc: 0.7331\n",
      "Epoch 52/200\n",
      "251/251 [==============================] - 0s 409us/sample - loss: 0.1054 - acc: 0.7171\n",
      "Epoch 53/200\n",
      "251/251 [==============================] - 0s 415us/sample - loss: 0.1025 - acc: 0.7012\n",
      "Epoch 54/200\n",
      "251/251 [==============================] - 0s 424us/sample - loss: 0.1086 - acc: 0.7012\n",
      "Epoch 55/200\n",
      "251/251 [==============================] - 0s 422us/sample - loss: 0.1093 - acc: 0.7251\n",
      "Epoch 56/200\n",
      "251/251 [==============================] - 0s 416us/sample - loss: 0.1004 - acc: 0.7490\n",
      "Epoch 57/200\n",
      "251/251 [==============================] - 0s 418us/sample - loss: 0.1011 - acc: 0.7291\n",
      "Epoch 58/200\n",
      "251/251 [==============================] - 0s 427us/sample - loss: 0.0990 - acc: 0.7371\n",
      "Epoch 59/200\n",
      "251/251 [==============================] - 0s 448us/sample - loss: 0.0948 - acc: 0.7410\n",
      "Epoch 60/200\n",
      "251/251 [==============================] - 0s 427us/sample - loss: 0.1048 - acc: 0.7012\n",
      "Epoch 61/200\n",
      "251/251 [==============================] - 0s 517us/sample - loss: 0.0981 - acc: 0.7410\n",
      "Epoch 62/200\n",
      "251/251 [==============================] - 0s 494us/sample - loss: 0.0949 - acc: 0.7610\n",
      "Epoch 63/200\n",
      "251/251 [==============================] - 0s 448us/sample - loss: 0.0982 - acc: 0.7331\n",
      "Epoch 64/200\n",
      "251/251 [==============================] - 0s 450us/sample - loss: 0.1082 - acc: 0.7171\n",
      "Epoch 65/200\n",
      "251/251 [==============================] - 0s 450us/sample - loss: 0.0999 - acc: 0.7410\n",
      "Epoch 66/200\n",
      "251/251 [==============================] - 0s 450us/sample - loss: 0.0951 - acc: 0.7490\n",
      "Epoch 67/200\n",
      "251/251 [==============================] - 0s 431us/sample - loss: 0.0939 - acc: 0.7610\n",
      "Epoch 68/200\n",
      "251/251 [==============================] - 0s 418us/sample - loss: 0.1009 - acc: 0.7610\n",
      "Epoch 69/200\n",
      "251/251 [==============================] - 0s 454us/sample - loss: 0.1031 - acc: 0.7291\n",
      "Epoch 70/200\n",
      "251/251 [==============================] - 0s 463us/sample - loss: 0.0993 - acc: 0.7649\n",
      "Epoch 71/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.1008 - acc: 0.7291\n",
      "Epoch 72/200\n",
      "251/251 [==============================] - 0s 479us/sample - loss: 0.0998 - acc: 0.7450\n",
      "Epoch 73/200\n",
      "251/251 [==============================] - 0s 449us/sample - loss: 0.0913 - acc: 0.7888\n",
      "Epoch 74/200\n",
      "251/251 [==============================] - 0s 462us/sample - loss: 0.0967 - acc: 0.7450\n",
      "Epoch 75/200\n",
      "251/251 [==============================] - 0s 483us/sample - loss: 0.0950 - acc: 0.7450\n",
      "Epoch 76/200\n",
      "251/251 [==============================] - 0s 512us/sample - loss: 0.0929 - acc: 0.7610\n",
      "Epoch 77/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.0903 - acc: 0.7769\n",
      "Epoch 78/200\n",
      "251/251 [==============================] - 0s 464us/sample - loss: 0.0953 - acc: 0.7450\n",
      "Epoch 79/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "251/251 [==============================] - 0s 431us/sample - loss: 0.0938 - acc: 0.7450\n",
      "Epoch 80/200\n",
      "251/251 [==============================] - 0s 415us/sample - loss: 0.0958 - acc: 0.7530\n",
      "Epoch 81/200\n",
      "251/251 [==============================] - 0s 453us/sample - loss: 0.0951 - acc: 0.7570\n",
      "Epoch 82/200\n",
      "251/251 [==============================] - 0s 449us/sample - loss: 0.0930 - acc: 0.7729\n",
      "Epoch 83/200\n",
      "251/251 [==============================] - 0s 427us/sample - loss: 0.0942 - acc: 0.7490\n",
      "Epoch 84/200\n",
      "251/251 [==============================] - 0s 426us/sample - loss: 0.0886 - acc: 0.7888\n",
      "Epoch 85/200\n",
      "251/251 [==============================] - 0s 438us/sample - loss: 0.0928 - acc: 0.7610\n",
      "Epoch 86/200\n",
      "251/251 [==============================] - 0s 449us/sample - loss: 0.0916 - acc: 0.7888\n",
      "Epoch 87/200\n",
      "251/251 [==============================] - 0s 426us/sample - loss: 0.0912 - acc: 0.7849\n",
      "Epoch 88/200\n",
      "251/251 [==============================] - 0s 425us/sample - loss: 0.0902 - acc: 0.7888\n",
      "Epoch 89/200\n",
      "251/251 [==============================] - 0s 441us/sample - loss: 0.0967 - acc: 0.7610\n",
      "Epoch 90/200\n",
      "251/251 [==============================] - 0s 439us/sample - loss: 0.0947 - acc: 0.7769\n",
      "Epoch 91/200\n",
      "251/251 [==============================] - 0s 422us/sample - loss: 0.0921 - acc: 0.7729\n",
      "Epoch 92/200\n",
      "251/251 [==============================] - 0s 421us/sample - loss: 0.0930 - acc: 0.7769\n",
      "Epoch 93/200\n",
      "251/251 [==============================] - 0s 461us/sample - loss: 0.0916 - acc: 0.7769\n",
      "Epoch 94/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.0897 - acc: 0.7769\n",
      "Epoch 95/200\n",
      "251/251 [==============================] - 0s 488us/sample - loss: 0.0933 - acc: 0.7729\n",
      "Epoch 96/200\n",
      "251/251 [==============================] - 0s 476us/sample - loss: 0.0907 - acc: 0.7649\n",
      "Epoch 97/200\n",
      "251/251 [==============================] - 0s 518us/sample - loss: 0.1000 - acc: 0.7450\n",
      "Epoch 98/200\n",
      "251/251 [==============================] - 0s 525us/sample - loss: 0.0932 - acc: 0.7769\n",
      "Epoch 99/200\n",
      "251/251 [==============================] - 0s 495us/sample - loss: 0.0880 - acc: 0.7968\n",
      "Epoch 100/200\n",
      "251/251 [==============================] - 0s 484us/sample - loss: 0.0883 - acc: 0.7928\n",
      "Epoch 101/200\n",
      "251/251 [==============================] - 0s 486us/sample - loss: 0.0972 - acc: 0.7450\n",
      "Epoch 102/200\n",
      "251/251 [==============================] - 0s 449us/sample - loss: 0.1078 - acc: 0.7570\n",
      "Epoch 103/200\n",
      "251/251 [==============================] - 0s 442us/sample - loss: 0.0980 - acc: 0.7729\n",
      "Epoch 104/200\n",
      "251/251 [==============================] - 0s 419us/sample - loss: 0.1053 - acc: 0.7530\n",
      "Epoch 105/200\n",
      "251/251 [==============================] - 0s 441us/sample - loss: 0.0947 - acc: 0.8008\n",
      "Epoch 106/200\n",
      "251/251 [==============================] - 0s 453us/sample - loss: 0.0899 - acc: 0.7888\n",
      "Epoch 107/200\n",
      "251/251 [==============================] - 0s 453us/sample - loss: 0.1051 - acc: 0.7291\n",
      "Epoch 108/200\n",
      "251/251 [==============================] - 0s 429us/sample - loss: 0.0878 - acc: 0.8048\n",
      "Epoch 109/200\n",
      "251/251 [==============================] - 0s 435us/sample - loss: 0.1124 - acc: 0.7331\n",
      "Epoch 110/200\n",
      "251/251 [==============================] - 0s 503us/sample - loss: 0.0921 - acc: 0.7888\n",
      "Epoch 111/200\n",
      "251/251 [==============================] - 0s 482us/sample - loss: 0.0910 - acc: 0.7769\n",
      "Epoch 112/200\n",
      "251/251 [==============================] - 0s 443us/sample - loss: 0.0881 - acc: 0.8008\n",
      "Epoch 113/200\n",
      "251/251 [==============================] - 0s 442us/sample - loss: 0.0859 - acc: 0.8088\n",
      "Epoch 114/200\n",
      "251/251 [==============================] - 0s 411us/sample - loss: 0.0839 - acc: 0.8287\n",
      "Epoch 115/200\n",
      "251/251 [==============================] - 0s 446us/sample - loss: 0.0975 - acc: 0.7251\n",
      "Epoch 116/200\n",
      "251/251 [==============================] - 0s 448us/sample - loss: 0.0984 - acc: 0.7729\n",
      "Epoch 117/200\n",
      "251/251 [==============================] - 0s 473us/sample - loss: 0.0876 - acc: 0.8048\n",
      "Epoch 118/200\n",
      "251/251 [==============================] - 0s 452us/sample - loss: 0.0872 - acc: 0.8008\n",
      "Epoch 119/200\n",
      "251/251 [==============================] - 0s 409us/sample - loss: 0.0847 - acc: 0.7968\n",
      "Epoch 120/200\n",
      "251/251 [==============================] - 0s 429us/sample - loss: 0.0857 - acc: 0.8127\n",
      "Epoch 121/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.0900 - acc: 0.7888\n",
      "Epoch 122/200\n",
      "251/251 [==============================] - 0s 461us/sample - loss: 0.0850 - acc: 0.8167\n",
      "Epoch 123/200\n",
      "251/251 [==============================] - 0s 465us/sample - loss: 0.0835 - acc: 0.8127\n",
      "Epoch 124/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.0829 - acc: 0.8247\n",
      "Epoch 125/200\n",
      "251/251 [==============================] - 0s 461us/sample - loss: 0.0844 - acc: 0.8367\n",
      "Epoch 126/200\n",
      "251/251 [==============================] - 0s 515us/sample - loss: 0.0990 - acc: 0.7530\n",
      "Epoch 127/200\n",
      "251/251 [==============================] - 0s 489us/sample - loss: 0.0873 - acc: 0.8127\n",
      "Epoch 128/200\n",
      "251/251 [==============================] - 0s 466us/sample - loss: 0.0910 - acc: 0.7769\n",
      "Epoch 129/200\n",
      "251/251 [==============================] - 0s 442us/sample - loss: 0.0912 - acc: 0.7968\n",
      "Epoch 130/200\n",
      "251/251 [==============================] - 0s 441us/sample - loss: 0.0894 - acc: 0.7809\n",
      "Epoch 131/200\n",
      "251/251 [==============================] - 0s 463us/sample - loss: 0.0873 - acc: 0.8088\n",
      "Epoch 132/200\n",
      "251/251 [==============================] - 0s 459us/sample - loss: 0.0815 - acc: 0.8088\n",
      "Epoch 133/200\n",
      "251/251 [==============================] - 0s 446us/sample - loss: 0.0902 - acc: 0.7809\n",
      "Epoch 134/200\n",
      "251/251 [==============================] - 0s 465us/sample - loss: 0.0995 - acc: 0.7689\n",
      "Epoch 135/200\n",
      "251/251 [==============================] - 0s 441us/sample - loss: 0.0856 - acc: 0.7928\n",
      "Epoch 136/200\n",
      "251/251 [==============================] - 0s 457us/sample - loss: 0.0836 - acc: 0.8127\n",
      "Epoch 137/200\n",
      "251/251 [==============================] - 0s 422us/sample - loss: 0.1043 - acc: 0.6813\n",
      "Epoch 138/200\n",
      "251/251 [==============================] - 0s 405us/sample - loss: 0.1005 - acc: 0.7490\n",
      "Epoch 139/200\n",
      "251/251 [==============================] - 0s 469us/sample - loss: 0.0903 - acc: 0.7888\n",
      "Epoch 140/200\n",
      "251/251 [==============================] - 0s 488us/sample - loss: 0.0866 - acc: 0.7928\n",
      "Epoch 141/200\n",
      "251/251 [==============================] - 0s 431us/sample - loss: 0.0841 - acc: 0.8008\n",
      "Epoch 142/200\n",
      "251/251 [==============================] - 0s 453us/sample - loss: 0.0831 - acc: 0.8088\n",
      "Epoch 143/200\n",
      "251/251 [==============================] - 0s 433us/sample - loss: 0.0800 - acc: 0.8367\n",
      "Epoch 144/200\n",
      "251/251 [==============================] - 0s 425us/sample - loss: 0.0798 - acc: 0.8167\n",
      "Epoch 145/200\n",
      "251/251 [==============================] - 0s 462us/sample - loss: 0.0790 - acc: 0.8406\n",
      "Epoch 146/200\n",
      "251/251 [==============================] - 0s 473us/sample - loss: 0.0856 - acc: 0.7849\n",
      "Epoch 147/200\n",
      "251/251 [==============================] - 0s 422us/sample - loss: 0.0824 - acc: 0.8008\n",
      "Epoch 148/200\n",
      "251/251 [==============================] - 0s 419us/sample - loss: 0.0860 - acc: 0.7968\n",
      "Epoch 149/200\n",
      "251/251 [==============================] - 0s 435us/sample - loss: 0.0850 - acc: 0.8048\n",
      "Epoch 150/200\n",
      "251/251 [==============================] - 0s 425us/sample - loss: 0.0833 - acc: 0.8048\n",
      "Epoch 151/200\n",
      "251/251 [==============================] - 0s 427us/sample - loss: 0.0810 - acc: 0.8287\n",
      "Epoch 152/200\n",
      "251/251 [==============================] - 0s 438us/sample - loss: 0.0856 - acc: 0.7928\n",
      "Epoch 153/200\n",
      "251/251 [==============================] - 0s 426us/sample - loss: 0.0822 - acc: 0.8207\n",
      "Epoch 154/200\n",
      "251/251 [==============================] - 0s 417us/sample - loss: 0.0855 - acc: 0.7968\n",
      "Epoch 155/200\n",
      "251/251 [==============================] - 0s 403us/sample - loss: 0.0890 - acc: 0.7968\n",
      "Epoch 156/200\n",
      "251/251 [==============================] - 0s 414us/sample - loss: 0.0897 - acc: 0.8008\n",
      "Epoch 157/200\n",
      "251/251 [==============================] - 0s 439us/sample - loss: 0.0854 - acc: 0.8167\n",
      "Epoch 158/200\n",
      "251/251 [==============================] - 0s 415us/sample - loss: 0.0839 - acc: 0.8167\n",
      "Epoch 159/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "251/251 [==============================] - 0s 442us/sample - loss: 0.0781 - acc: 0.8247\n",
      "Epoch 160/200\n",
      "251/251 [==============================] - 0s 427us/sample - loss: 0.0894 - acc: 0.7649\n",
      "Epoch 161/200\n",
      "251/251 [==============================] - 0s 410us/sample - loss: 0.0823 - acc: 0.8167\n",
      "Epoch 162/200\n",
      "251/251 [==============================] - 0s 413us/sample - loss: 0.0849 - acc: 0.8167\n",
      "Epoch 163/200\n",
      "251/251 [==============================] - 0s 412us/sample - loss: 0.0830 - acc: 0.8207\n",
      "Epoch 164/200\n",
      "251/251 [==============================] - 0s 413us/sample - loss: 0.0839 - acc: 0.8327\n",
      "Epoch 165/200\n",
      "251/251 [==============================] - 0s 400us/sample - loss: 0.0790 - acc: 0.8367\n",
      "Epoch 166/200\n",
      "251/251 [==============================] - 0s 412us/sample - loss: 0.0774 - acc: 0.8446\n",
      "Epoch 167/200\n",
      "251/251 [==============================] - 0s 412us/sample - loss: 0.1087 - acc: 0.6972\n",
      "Epoch 168/200\n",
      "251/251 [==============================] - 0s 414us/sample - loss: 0.0889 - acc: 0.8167\n",
      "Epoch 169/200\n",
      "251/251 [==============================] - 0s 421us/sample - loss: 0.0832 - acc: 0.8088\n",
      "Epoch 170/200\n",
      "251/251 [==============================] - 0s 419us/sample - loss: 0.0800 - acc: 0.8207\n",
      "Epoch 171/200\n",
      "251/251 [==============================] - 0s 414us/sample - loss: 0.0807 - acc: 0.8367\n",
      "Epoch 172/200\n",
      "251/251 [==============================] - 0s 415us/sample - loss: 0.0801 - acc: 0.8247\n",
      "Epoch 173/200\n",
      "251/251 [==============================] - 0s 417us/sample - loss: 0.0800 - acc: 0.8287\n",
      "Epoch 174/200\n",
      "251/251 [==============================] - 0s 416us/sample - loss: 0.0817 - acc: 0.8167\n",
      "Epoch 175/200\n",
      "251/251 [==============================] - 0s 424us/sample - loss: 0.0978 - acc: 0.7570\n",
      "Epoch 176/200\n",
      "251/251 [==============================] - 0s 416us/sample - loss: 0.0914 - acc: 0.8008\n",
      "Epoch 177/200\n",
      "251/251 [==============================] - 0s 422us/sample - loss: 0.0831 - acc: 0.8207\n",
      "Epoch 178/200\n",
      "251/251 [==============================] - 0s 422us/sample - loss: 0.0813 - acc: 0.8327\n",
      "Epoch 179/200\n",
      "251/251 [==============================] - 0s 415us/sample - loss: 0.0797 - acc: 0.8247\n",
      "Epoch 180/200\n",
      "251/251 [==============================] - 0s 423us/sample - loss: 0.0795 - acc: 0.8367\n",
      "Epoch 181/200\n",
      "251/251 [==============================] - 0s 409us/sample - loss: 0.0796 - acc: 0.8367\n",
      "Epoch 182/200\n",
      "251/251 [==============================] - 0s 415us/sample - loss: 0.0759 - acc: 0.8486\n",
      "Epoch 183/200\n",
      "251/251 [==============================] - 0s 413us/sample - loss: 0.0787 - acc: 0.8127\n",
      "Epoch 184/200\n",
      "251/251 [==============================] - 0s 401us/sample - loss: 0.0752 - acc: 0.8486\n",
      "Epoch 185/200\n",
      "251/251 [==============================] - 0s 411us/sample - loss: 0.0742 - acc: 0.8526\n",
      "Epoch 186/200\n",
      "251/251 [==============================] - 0s 401us/sample - loss: 0.0789 - acc: 0.8207\n",
      "Epoch 187/200\n",
      "251/251 [==============================] - 0s 409us/sample - loss: 0.0751 - acc: 0.8247\n",
      "Epoch 188/200\n",
      "251/251 [==============================] - 0s 402us/sample - loss: 0.0757 - acc: 0.8367\n",
      "Epoch 189/200\n",
      "251/251 [==============================] - 0s 403us/sample - loss: 0.0870 - acc: 0.7769\n",
      "Epoch 190/200\n",
      "251/251 [==============================] - 0s 408us/sample - loss: 0.0789 - acc: 0.8406\n",
      "Epoch 191/200\n",
      "251/251 [==============================] - 0s 417us/sample - loss: 0.0789 - acc: 0.8247\n",
      "Epoch 192/200\n",
      "251/251 [==============================] - 0s 411us/sample - loss: 0.0845 - acc: 0.7968\n",
      "Epoch 193/200\n",
      "251/251 [==============================] - 0s 411us/sample - loss: 0.0760 - acc: 0.8367\n",
      "Epoch 194/200\n",
      "251/251 [==============================] - 0s 412us/sample - loss: 0.0768 - acc: 0.8247\n",
      "Epoch 195/200\n",
      "251/251 [==============================] - 0s 400us/sample - loss: 0.0781 - acc: 0.8287\n",
      "Epoch 196/200\n",
      "251/251 [==============================] - 0s 410us/sample - loss: 0.0800 - acc: 0.8008\n",
      "Epoch 197/200\n",
      "251/251 [==============================] - 0s 408us/sample - loss: 0.0886 - acc: 0.8088\n",
      "Epoch 198/200\n",
      "251/251 [==============================] - 0s 416us/sample - loss: 0.0838 - acc: 0.8247\n",
      "Epoch 199/200\n",
      "251/251 [==============================] - 0s 406us/sample - loss: 0.0746 - acc: 0.8526\n",
      "Epoch 200/200\n",
      "251/251 [==============================] - 0s 407us/sample - loss: 0.0740 - acc: 0.8367\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(train_x, train_y, epochs=200, batch_size=10, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "63/63 [==============================] - 0s 1ms/sample - loss: 0.0910 - acc: 0.7937\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.09099126689963871, 0.7936508]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(test_x, test_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.07232401, 0.805886  , 0.02867882, 0.05921362, 0.03389756]],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.predict(test_x[10].reshape(1, 1, 270))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save(\"gesture_lstm_v10.h5\")"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}