offline_test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os,sys,time\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import librosa\n",
    "sys.path.insert(0, './scripts')\n",
    "sys.path.insert(0, './models')\n",
    "import e2e_model as nn_model_foreval\n",
    "\n",
    "def cmvn_slide(feat,winlen=300,cmvn=False): #feat : (length, dim) 2d matrix\n",
    "# function for Cepstral Mean Variance Nomalization\n",
    "\n",
    "    maxlen = np.shape(feat)[0]\n",
    "    new_feat = np.empty_like(feat)\n",
    "    cur = 1\n",
    "    leftwin = 0\n",
    "    rightwin = winlen/2\n",
    "    \n",
    "    # middle\n",
    "    for cur in range(maxlen):\n",
    "        cur_slide = feat[cur-leftwin:cur+rightwin,:] \n",
    "        #cur_slide = feat[cur-winlen/2:cur+winlen/2,:]\n",
    "        mean =np.mean(cur_slide,axis=0)\n",
    "        std = np.std(cur_slide,axis=0)\n",
    "        if cmvn == 'mv':\n",
    "            new_feat[cur,:] = (feat[cur,:]-mean)/std # for cmvn        \n",
    "        elif cmvn =='m':\n",
    "            new_feat[cur,:] = (feat[cur,:]-mean) # for cmn\n",
    "        if leftwin<winlen/2:\n",
    "            leftwin+=1\n",
    "        elif maxlen-cur < winlen/2:\n",
    "            rightwin-=1    \n",
    "    return new_feat\n",
    "\n",
    "\n",
    "def feat_extract(filelist,feat_type,n_fft_length=512,hop=160,vad=True,cmvn=False,exclude_short=500):\n",
    "# function for feature extracting\n",
    "        \n",
    "    feat = []\n",
    "    utt_shape = []\n",
    "    new_utt_label =[]\n",
    "    for index,wavname in enumerate(filelist):\n",
    "        #read audio input\n",
    "        y, sr = librosa.core.load(wavname,sr=16000,mono=True,dtype='float')\n",
    "\n",
    "        #extract feature\n",
    "        if feat_type =='melspec':\n",
    "            Y = librosa.feature.melspectrogram(y,sr,n_fft=n_fft_length,hop_length=hop,n_mels=40,fmin=133,fmax=6955)\n",
    "        elif feat_type =='mfcc':\n",
    "            Y = librosa.feature.mfcc(y,sr,n_fft=n_fft_length,hop_length=hop,n_mfcc=40,fmin=133,fmax=6955)\n",
    "        elif feat_type =='spec':\n",
    "            Y = np.abs( librosa.core.stft(y,n_fft=n_fft_length,hop_length=hop,win_length=400) )\n",
    "        elif feat_type =='logspec':\n",
    "            Y = np.log( np.abs( librosa.core.stft(y,n_fft=n_fft_length,hop_length=hop,win_length=400) ) )\n",
    "        elif feat_type =='logmel':\n",
    "            Y = np.log( librosa.feature.melspectrogram(y,sr,n_fft=n_fft_length,hop_length=hop,n_mels=40,fmin=133,fmax=6955) )\n",
    "\n",
    "        Y = Y.transpose()\n",
    "        \n",
    "        \n",
    "        # Simple VAD based on the energy\n",
    "        if vad:\n",
    "            E = librosa.feature.rmse(y, frame_length=n_fft_length,hop_length=hop,)\n",
    "            threshold= np.mean(E)/2 * 1.04\n",
    "            vad_segments = np.nonzero(E>threshold)\n",
    "            if vad_segments[1].size!=0:\n",
    "                Y = Y[vad_segments[1],:]\n",
    "\n",
    "                \n",
    "        #exclude short utterance under \"exclude_short\" value\n",
    "        if exclude_short == 0 or (Y.shape[0] > exclude_short):\n",
    "            if cmvn:\n",
    "                Y = cmvn_slide(Y,300,cmvn)\n",
    "            feat.append(Y)\n",
    "            utt_shape.append(np.array(Y.shape))\n",
    "#             new_utt_label.append(utt_label[index])\n",
    "            sys.stdout.write('%s\\r' % index)\n",
    "            sys.stdout.flush()\n",
    "            \n",
    "        if index ==0:\n",
    "            break\n",
    "\n",
    "        \n",
    "    tffilename = feat_type+'_fft'+str(n_fft_length)+'_hop'+str(hop)\n",
    "    if vad:\n",
    "        tffilename += '_vad'\n",
    "    if cmvn=='m':\n",
    "        tffilename += '_cmn'\n",
    "    elif cmvn =='mv':\n",
    "        tffilename += '_cmvn'\n",
    "    if exclude_short >0:\n",
    "        tffilename += '_exshort'+str(exclude_short)\n",
    "\n",
    "    return feat, new_utt_label, utt_shape, tffilename #feat : (length, dim) 2d matrix\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.224723815918 seconds\n",
      "(1, 914, 40)\n"
     ]
    }
   ],
   "source": [
    "# Feature extraction configuration\n",
    "FEAT_TYPE = 'logmel'\n",
    "N_FFT = 400\n",
    "HOP = 160\n",
    "VAD = True\n",
    "CMVN = 'mv'\n",
    "EXCLUDE_SHORT=0\n",
    "IS_BATCHNORM = False\n",
    "IS_TRAINING = False\n",
    "INPUT_DIM = 40\n",
    "\n",
    "# extracting mfcc for input wavfile\n",
    "FILENAME = ['/data/sls/qcri/asr/data/vardial/vardial2017/test.vardial2017/wav/EGY/04d3ad10aceb69fcfb3a55d102ba7cff_M_0004_116.66_129.74.wav']\n",
    "start_time = time.time()\n",
    "\n",
    "feat, utt_label, utt_shape, tffilename = feat_extract(FILENAME,FEAT_TYPE,N_FFT,HOP,VAD,CMVN,EXCLUDE_SHORT)\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print format(elapsed_time) + ' seconds'\n",
    "print np.shape(feat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/data/sls/u/swshon/tools/pytf/local/lib/python2.7/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"transpose:0\", shape=(?, ?, ?), dtype=float32)\n",
      "Tensor(\"transpose_1:0\", shape=(?, ?, ?), dtype=float32)\n",
      "Tensor(\"transpose_2:0\", shape=(?, ?, ?), dtype=float32)\n",
      "Tensor(\"transpose_3:0\", shape=(?, ?, ?), dtype=float32)\n",
      "Tensor(\"Mul:0\", shape=(?, ?, 500), dtype=float32)\n",
      "(?, 3000)\n",
      "(?, 1500)\n",
      "(?, 600)\n",
      "WARNING:tensorflow:From /data/sls/u/swshon/tools/pytf/local/lib/python2.7/site-packages/tensorflow/python/util/tf_should_use.py:175: initialize_all_variables (from tensorflow.python.ops.variables) is deprecated and will be removed after 2017-03-02.\n",
      "Instructions for updating:\n",
      "Use `tf.global_variables_initializer` instead.\n",
      "INFO:tensorflow:Restoring parameters from ./data/pretrained_model/model1284000.ckpt-1284000\n"
     ]
    }
   ],
   "source": [
    "# Variable Initialization    \n",
    "\n",
    "softmax_num = 5\n",
    "x = tf.placeholder(tf.float32, [None,None,40])\n",
    "y = tf.placeholder(tf.int32, [None])\n",
    "s = tf.placeholder(tf.int32, [None,2])\n",
    "\n",
    "emnet_validation = nn_model_foreval.nn(x,y,y,s,softmax_num,IS_TRAINING,INPUT_DIM,IS_BATCHNORM);\n",
    "sess = tf.InteractiveSession()\n",
    "saver = tf.train.Saver()\n",
    "tf.initialize_all_variables().run()\n",
    "\n",
    "### Loading neural network \n",
    "saver.restore(sess,'./data/pretrained_model/model1284000.ckpt-1284000')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.9759850502 seconds\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "likelihood= emnet_validation.o1.eval({x:feat, s:utt_shape})\n",
    "\n",
    "elapsed_time = time.time() - start_time\n",
    "print format(elapsed_time) + ' seconds'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The Input wav file /data/sls/qcri/asr/data/vardial/vardial2017/test.vardial2017/wav/EGY/04d3ad10aceb69fcfb3a55d102ba7cff_M_0004_116.66_129.74.wav is\n",
      " [Egytion]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,u'Dialect identification offline test result')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAEICAYAAABfz4NwAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZoATgpmZZf8LjRJlYqJf1egAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f23f87798d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Print out the results in barplot\n",
    "# print likelihood\n",
    "dialect_index = np.argmax(likelihood)\n",
    "\n",
    "print 'The Input wav file '+FILENAME[0]+' is'\n",
    "if dialect_index == 0:\n",
    "    print ' [Egytion]'\n",
    "elif dialect_index == 1:\n",
    "    print ' [Gulf]'\n",
    "elif dialect_index == 2:\n",
    "    print ' [Levantine]'\n",
    "elif dialect_index == 3:\n",
    "    print ' [Modern Standard Arabic]'\n",
    "elif dialect_index == 4:\n",
    "    print ' [North African]'\n",
    "    \n",
    "\n",
    "import matplotlib.pyplot as plt; plt.rcdefaults()\n",
    "import matplotlib.pyplot as plt\n",
    "# matplotlib.use('Agg')\n",
    "%matplotlib inline\n",
    "\n",
    "x = np.arange(5)\n",
    "dialects = ['Egytion','Gulf','Levantine','MSA','North Africa']\n",
    "plt.bar(x,likelihood[0],align='center')\n",
    "plt.xticks(x,dialects)\n",
    "plt.ylabel('Likelihood')\n",
    "plt.title('Dialect identification offline test result')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}