USC_CNN_v2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Urban Sound Classifier using CNN v2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "第一个CNN的分类器效果惨烈，有许多细节都没有注意到，感觉修改也令人心烦，从头开始。老实说我也不知道这次能不能成功，边做边看吧。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 上次踩过的坑和经验"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. 之前把频谱当作训练数据输入，但频谱里的数值都是负数，后来换为了绝对值。但这样其实也不行，应该要进行normalization。把是把值的范围限制在[-1,1]还是[0,1]还不清楚。我的想法是根据relu的性质，应当是[0,1]。（结果：应当时-1到1，值是*w+b再激活，所以负数不影响）\n",
    "2. 计算mfcc时的维度一般是40，有用的一般是2到13维。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 构想"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "首先明确一下这次要做的事情：\n",
    "\n",
    "- 训练集：[UrbanSound8K dataset](https://serv.cusp.nyu.edu/projects/urbansounddataset/urbansound8k.html)，训练用fold 1-3，validation用从整个数据集里面随意找一些样本，test用fold 10。\n",
    "- 输入：MFCC，normalization为mean=0, var=1。\n",
    "- 网络：一个非常简单的CNN，两层conv2D层，进入一个全连接层。\n",
    "- 输出：10个class"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 动手，去做"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'2.2.0'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# draw\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('ggplot')\n",
    "# basic handling\n",
    "import os\n",
    "import glob\n",
    "import pickle\n",
    "import numpy as np\n",
    "# audio\n",
    "import librosa\n",
    "import librosa.display\n",
    "import IPython.display\n",
    "# normalization\n",
    "import sklearn\n",
    "# nn\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Convolution2D, MaxPooling2D\n",
    "from keras.utils import to_categorical\n",
    "from keras.callbacks import LearningRateScheduler\n",
    "keras.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "指定dataset的位置"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "parent_dir = '../data/UrbanSound8K/audio/'\n",
    "\n",
    "train_dir = 'train/'\n",
    "val_dir = 'val/'\n",
    "test_dir = 'fold10/'\n",
    "\n",
    "file_name = '*.wav'\n",
    "\n",
    "train_files = glob.glob(os.path.join(parent_dir, train_dir, file_name))\n",
    "val_files = glob.glob(os.path.join(parent_dir, val_dir, file_name))\n",
    "test_files = glob.glob(os.path.join(parent_dir, test_dir, file_name))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "定义一个函数用于读取音频片段，库里的片段几乎都是4s，但有一部分小于4秒，将它们补零。采样率22050，4秒一共88200个采样点。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_clip(filename):\n",
    "    x, sr = librosa.load(filename)\n",
    "    x = np.pad(x,(0,88200-x.shape[0]),'constant')\n",
    "    return x, sr"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "再定义一个函数，用于提取片段的mfcc并进行normalization。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def extract_feature(filename):\n",
    "    x, sr = load_clip(filename)\n",
    "    mfccs = librosa.feature.mfcc(y=x, sr=sr, n_mfcc=40)\n",
    "    norm_mfccs = sklearn.preprocessing.scale(mfccs, axis=1)\n",
    "    return norm_mfccs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]\n",
      "[-1.25692648e-14  1.57869864e-16  5.60887239e-16  4.24836787e-16\n",
      "  1.40542683e-16  6.23778486e-16  4.52431926e-17  4.46014452e-17\n",
      " -4.88530218e-17 -4.99921235e-16 -3.52640203e-16 -4.04300870e-17\n",
      " -6.48164887e-17 -5.00562982e-17 -1.41505305e-16  1.22573756e-16\n",
      "  1.66854327e-17 -4.73609591e-16 -1.86748497e-16 -4.96712498e-16\n",
      " -7.89349318e-17 -2.19477615e-16 -3.76063984e-16 -8.41972606e-16\n",
      "  0.00000000e+00 -2.95203810e-16 -3.54244572e-16  2.27820332e-16\n",
      "  2.80443620e-16 -2.25253342e-16 -2.65041682e-16 -4.06867860e-16\n",
      " -1.21932008e-16  3.97883396e-17  5.10830941e-16  1.84823255e-16\n",
      "  3.38842634e-16 -8.98446378e-18  3.70930005e-16 -1.42147052e-16]\n"
     ]
    }
   ],
   "source": [
    "# 测试extract_feature是否正常工作\n",
    "mfccs = extract_feature('./1.wav')\n",
    "plt.figure(figsize=(20,5))\n",
    "librosa.display.specshow(mfccs, sr=22050, x_axis='time', cmap='viridis')\n",
    "plt.colorbar()\n",
    "plt.show()\n",
    "print (mfccs.var(axis=1))\n",
    "print (mfccs.mean(axis=1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "读取整个数据集，从整个数据集提取特征与标签"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_dataset(filenames):\n",
    "    features, labels = np.empty((0,40,173)), np.empty(0)\n",
    "    cnt = 0;\n",
    "    cnt_all = len(filenames)\n",
    "    \n",
    "    for filename in filenames:\n",
    "        mfccs = extract_feature(filename)\n",
    "        features = np.append(features,mfccs[None],axis=0)\n",
    "        cnt+=1\n",
    "        if(cnt%100==0):\n",
    "            print([str(cnt)+' / '+str(cnt_all)+' finished'])\n",
    "        labels = np.append(labels, filename.split('\\\\')[1].split('-')[1])\n",
    "    return np.array(features), np.array(labels, dtype=np.int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "将训练过程可视化的函数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show_history(history):\n",
    "    print(history.history.keys())\n",
    "    fig = plt.figure(figsize=(20,5))\n",
    "    plt.subplot(121)\n",
    "    plt.plot(history.history['acc'])\n",
    "    plt.plot(history.history['val_acc'])\n",
    "    plt.title('model accuracy')\n",
    "    plt.ylabel('accuracy')\n",
    "    plt.xlabel('epoch')\n",
    "    plt.legend(['train', 'test'], loc='upper left')\n",
    "    plt.subplot(122)\n",
    "    plt.plot(history.history['loss'])\n",
    "    plt.plot(history.history['val_loss'])\n",
    "    plt.title('model loss')\n",
    "    plt.ylabel('loss')\n",
    "    plt.xlabel('epoch')\n",
    "    plt.legend(['train', 'test'], loc='lower left')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "如果还没有将音频转换为features，则进行转化并保存。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train_x, train_y = load_dataset(train_files)\n",
    "# pickle.dump(train_x, open('./train_x.dat', 'wb'))\n",
    "# pickle.dump(train_y, open('./train_y.dat', 'wb'))\n",
    "\n",
    "# val_x, val_y = load_dataset(val_files)\n",
    "# pickle.dump(val_x, open('./val_x.dat', 'wb'))\n",
    "# pickle.dump(val_y, open('./val_y.dat', 'wb'))\n",
    "\n",
    "# test_x, test_y = load_dataset(test_files)\n",
    "# pickle.dump(test_x, open('./test_x.dat', 'wb'))\n",
    "# pickle.dump(test_y, open('./test_y.dat', 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "如果已经有features了，就从文件中读取。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x = pickle.load(open('./train_x.dat', 'rb'))\n",
    "train_y = pickle.load(open('./train_y.dat', 'rb'))\n",
    "val_x = pickle.load(open('./val_x.dat', 'rb'))\n",
    "val_y = pickle.load(open('./val_y.dat', 'rb'))\n",
    "test_x = pickle.load(open('./test_x.dat', 'rb'))\n",
    "test_y = pickle.load(open('./test_y.dat', 'rb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "接下来对feature进行一些预处理。mfcc是二维数据，要输入conv2D层，要把它变为3维数据。因为是聚类，所以把label转化为categories"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x = train_x.reshape(train_x.shape[0],train_x.shape[1],train_x.shape[2],1)\n",
    "val_x = val_x.reshape(val_x.shape[0],val_x.shape[1],val_x.shape[2],1)\n",
    "test_x = test_x.reshape(test_x.shape[0],test_x.shape[1],test_x.shape[2],1)\n",
    "\n",
    "train_y = to_categorical(train_y)\n",
    "val_y = to_categorical(val_y)\n",
    "test_y = to_categorical(test_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pre-processing至此结束，检查一下我们的训练集"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2686, 40, 173, 1)\n",
      "(2686, 10)\n"
     ]
    }
   ],
   "source": [
    "print(train_x.shape)\n",
    "print(train_y.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "没有问题的话就开始搭建模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "Layer (type)                        Output Shape                    Param #     \n",
      "================================================================================\n",
      "conv2d_1 (Conv2D)                   (None, 38, 171, 32)             320         \n",
      "________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)      (None, 19, 85, 32)              0           \n",
      "________________________________________________________________________________\n",
      "dropout_1 (Dropout)                 (None, 19, 85, 32)              0           \n",
      "________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)                   (None, 17, 83, 32)              9248        \n",
      "________________________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2D)      (None, 8, 41, 32)               0           \n",
      "________________________________________________________________________________\n",
      "dropout_2 (Dropout)                 (None, 8, 41, 32)               0           \n",
      "________________________________________________________________________________\n",
      "flatten_1 (Flatten)                 (None, 10496)                   0           \n",
      "________________________________________________________________________________\n",
      "dense_1 (Dense)                     (None, 10)                      104970      \n",
      "================================================================================\n",
      "Total params: 114,538\n",
      "Trainable params: 114,538\n",
      "Non-trainable params: 0\n",
      "________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Convolution2D(32, (3, 3), activation='relu',input_shape = train_x.shape[1:]))\n",
    "model.add(MaxPooling2D((2, 2)))\n",
    "model.add(Dropout(0.5)) \n",
    "model.add(Convolution2D(32, (3, 3),  activation='relu'))\n",
    "model.add(MaxPooling2D(2, 2))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(10, activation='softmax'))\n",
    "model.compile(optimizer='Adam',\n",
    "                loss='categorical_crossentropy',\n",
    "                metrics=['accuracy'])\n",
    "model.summary(line_length=80)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 2686 samples, validate on 83 samples\n",
      "Epoch 1/10\n",
      "2686/2686 [==============================] - 13s 5ms/step - loss: 2.0973 - acc: 0.2223 - val_loss: 1.8910 - val_acc: 0.3012\n",
      "Epoch 2/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 1.5507 - acc: 0.4151 - val_loss: 1.6829 - val_acc: 0.3494\n",
      "Epoch 3/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 1.2991 - acc: 0.5316 - val_loss: 1.4978 - val_acc: 0.4458\n",
      "Epoch 4/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 1.1322 - acc: 0.5934 - val_loss: 1.7221 - val_acc: 0.3976\n",
      "Epoch 5/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 1.0006 - acc: 0.6385 - val_loss: 1.7749 - val_acc: 0.4096\n",
      "Epoch 6/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 0.9274 - acc: 0.6660 - val_loss: 1.4768 - val_acc: 0.4578\n",
      "Epoch 7/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 0.7824 - acc: 0.7226 - val_loss: 1.4234 - val_acc: 0.4819\n",
      "Epoch 8/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 0.7153 - acc: 0.7390 - val_loss: 1.5800 - val_acc: 0.4699\n",
      "Epoch 9/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 0.6537 - acc: 0.7640 - val_loss: 1.6162 - val_acc: 0.5301\n",
      "Epoch 10/10\n",
      "2686/2686 [==============================] - 8s 3ms/step - loss: 0.5779 - acc: 0.7897 - val_loss: 1.3760 - val_acc: 0.5060\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(train_x, train_y, epochs=10, batch_size=32, validation_data=(val_x, val_y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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 1440x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_history(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (tensorflow)",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}