Testing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL_DIR=\"model\"\n",
    "DATA_DIR=\"data\"\n",
    "SPECTROGRAM_TRAIN_TFRECORD_FILENAME = \"spectrogram_train.tfrecord\"\n",
    "SPECTROGRAM_VALID_TFRECORD_FILENAME = \"spectrogram_valid.tfrecord\"\n",
    "SPECTROGRAM_TEST_TFRECORD_FILENAME = \"spectrogram_test.tfrecord\"\n",
    "\n",
    "BUFFER_SIZE = 64\n",
    "BATCH_SIZE = 16\n",
    "SPECTROGRAM_SHAPE = (85, 129)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.models.load_model('model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_dataset = tf.data.TFRecordDataset([os.path.join(DATA_DIR, SPECTROGRAM_TEST_TFRECORD_FILENAME)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_description = {\n",
    "    'spectrogram': tf.io.FixedLenFeature([], tf.string, default_value=''),\n",
    "    'label': tf.io.FixedLenFeature([], tf.float32, default_value=0.0)\n",
    "}\n",
    "\n",
    "def _parse_function(example_proto):\n",
    "    example = tf.io.parse_single_example(example_proto, feature_description)\n",
    "    spectrogram = tf.io.parse_tensor(example['spectrogram'], out_type=tf.float32)\n",
    "    spectrogram = tf.ensure_shape(spectrogram, SPECTROGRAM_SHAPE)\n",
    "    label = tf.ensure_shape(example['label'], ())\n",
    "    return spectrogram, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_dataset = test_dataset.map(_parse_function)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "testing_dataset = test_dataset.batch(BATCH_SIZE).prefetch(BUFFER_SIZE).cache()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Run the model on CPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "93/93 [==============================] - 23s 243ms/step - loss: 0.0278 - accuracy: 0.9906\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.02776455320417881, 0.9905533194541931]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(testing_dataset)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since 1 batch contains 4 seconds of data, this is sufficient."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Precision Recall analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = np.argmax(model.predict(testing_dataset), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = np.array([label for spectrogram, label in test_dataset.as_numpy_iterator()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9933775"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m = tf.keras.metrics.Precision()\n",
    "m.update_state(y, y_pred)\n",
    "m.result().numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.96153843"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m = tf.keras.metrics.Recall()\n",
    "m.update_state(y, y_pred)\n",
    "m.result().numpy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The model achieves sufficiently high precision and recall on the test set."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test on actual audio"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Taken from here: https://www.youtube.com/watch?v=mxnL7UrkmY4 \n",
    "\n",
    "Note that this audio is noisier than the training data. It contains extra factors including narration and speech. These factors were not accounted for in the training data because the algorithm's purpose is not to classify applauses from other noise."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "rach, rach_sampling_rate = tf.audio.decode_wav(tf.io.read_file(os.path.join(DATA_DIR, 'rachmaninoff.wav')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: shape=(), dtype=int32, numpy=48000>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rach_sampling_rate"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since the sampling rate is different from our training set, we need to make the splits shorter. (The split lengths can't be changed from 11025 for STFT.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def split_data(d, d_sampling_rate):\n",
    "    data_size = d.shape[0]\n",
    "    batch_size = 44100 // 4\n",
    "    batch_count = data_size // batch_size \n",
    "    return np.array_split(d[:batch_count * batch_size], batch_count)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "rach_split = split_data(rach, rach_sampling_rate)\n",
    "DATA_SIZE = len(rach_split)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x12313c2d460>,\n",
       " <matplotlib.lines.Line2D at 0x12313c2d670>]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(20, 5))\n",
    "plt.plot(rach, color='blue')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = tf.data.Dataset.from_tensor_slices(rach_split)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_spectrogram(waveform):\n",
    "    spectrogram1 = tf.signal.stft(waveform[:, 0], frame_length=256, frame_step=128)\n",
    "    spectrogram1 = tf.abs(spectrogram1)\n",
    "    return spectrogram1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:AutoGraph could not transform <function get_spectrogram at 0x000001231AA27700> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: module 'gast' has no attribute 'Index'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function get_spectrogram at 0x000001231AA27700> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: module 'gast' has no attribute 'Index'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
     ]
    }
   ],
   "source": [
    "dataset = dataset.map(get_spectrogram)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = dataset.batch(BATCH_SIZE).prefetch(BUFFER_SIZE).cache()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model.predict(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = np.argmax(y_pred, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1231bf44370>]"
      ]
     },
     "execution_count": 55,
     "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.scatter(np.arange(y_pred.shape[0]), y_pred, s=1)\n",
    "plt.plot(np.linspace(51, 51), np.linspace(0, 1), c='red')\n",
    "plt.plot(np.linspace(70, 70), np.linspace(0, 1), c='green')\n",
    "plt.plot(np.linspace(960, 960), np.linspace(0, 1), c='red')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The video contains applauses roughly from 0:00-0:11 (0-51) and 3:41-4:17 (960-1122), and music from 0:16-3:40 (70-959). (Boundaries plotted above) \n",
    "\n",
    "It looks like the model predicts the labels quite conservatively, where precision seems to be very high (from inspection) but recall is not. This might be because in the applauses at the end also involve speech from the narrator. Also note that the time sections which are neither applauses or music (the performer is speaking/waiting and the audience is quiet) is classified negatively."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sequential cleanup"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "We would like to incorporate sequential data, as much as the circumstances allow. The algorithm would likely have a harder time deciphering the boundary points if the data flows one way (i.e. in an RNN). Instead, we will try a simpler algorithm - if a single time section of one label is bounded on both sides by a time sections of a different label, and either one of the bounding time sections is consecutively connected, the former time section is to be converted. \n",
    "\n",
    "As an example:\n",
    "\n",
    "1 0 1 1 -> 1 1 1 1\n",
    "0 1 0 1 -> 0 1 0 1\n",
    "1 1 0 1 0 1 0 0 -> 1 1 1 1 0 0 0 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "def transform(s):\n",
    "    t = np.zeros(s.shape)\n",
    "    t[0] = s[0]\n",
    "    for i in range(1, s.shape[0] - 1):\n",
    "        if s[i - 1] != s[i] and s[i + 1] != s[i]:\n",
    "            if (i - 2 >= 0 and s[i - 2] != s[i]) or (i + 2 < s.shape[0] and s[i + 2] != s[i]):\n",
    "                t[i] = s[i - 1]\n",
    "            else:\n",
    "                t[i] = s[i]\n",
    "        else:\n",
    "            t[i] = s[i]\n",
    "    t[s.shape[0] - 1] = s[s.shape[0] - 1]\n",
    "    return t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_trans = transform(y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1231bee1850>]"
      ]
     },
     "execution_count": 54,
     "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.scatter(np.arange(y_trans.shape[0]), y_trans, s=1)\n",
    "plt.plot(np.linspace(51, 51), np.linspace(0, 1), c='red')\n",
    "plt.plot(np.linspace(70, 70), np.linspace(0, 1), c='green')\n",
    "plt.plot(np.linspace(960, 960), np.linspace(0, 1), c='red')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "351\n",
      "355\n",
      "356\n",
      "484\n",
      "485\n"
     ]
    }
   ],
   "source": [
    "for i in range(300, 500):\n",
    "    if y_pred[i] == 1:\n",
    "        print(i)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Notice that this process cannot be satisfied with a trivial algorithm. Due to the nature of imperfectness of models, it seems difficult to achieve perfection. \n",
    "\n",
    "The incorporation of sequential data will be a topic for another study. We predict that it will be a similar task to video segment classification using an RNN."
   ]
  },
  {
   "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}