du16_test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "from tensorflow.keras.models import model_from_json\n",
    "from pystoi.stoi import stoi\n",
    "import mir_eval #https://github.com/craffel/mir_eval\n",
    "import librosa\n",
    "import tensorflow as tf\n",
    "# import keras\n",
    "from scipy.io import wavfile\n",
    "# from pypesq import pesq #https://github.com/vBaiCai/python-pesq\n",
    "# import vqmetrics\n",
    "from natsort import natsorted\n",
    "from tensorflow.keras.models import load_model\n",
    "from sklearn import preprocessing\n",
    "import matplotlib.pyplot as plt\n",
    "import pylab \n",
    "import soundfile as sf\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#define reconstruct function to reconstruct sound from framed signal.\n",
    "def reconstruct(wave,angle):\n",
    "    recon = np.sqrt(np.power(10, wave))\n",
    "    recon1 = recon*np.cos(angle)+recon*np.sin(angle)*1j\n",
    "    recon = librosa.core.istft((recon1.T), hop_length=256, win_length=512, window='hann')\n",
    "#     print(type(recon[0]))\n",
    "    return recon\n",
    "def change_order(first):\n",
    "    sec=np.copy(first)\n",
    "    sec=np.copy(first)\n",
    "    sec[0] = first[5]\n",
    "    sec[1] = first[4]\n",
    "    sec[2] = first[3]\n",
    "    sec[5] = first[2]\n",
    "    sec[4] = first[1]\n",
    "    sec[3] = first[0]\n",
    "    return sec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\ASUS\\Anaconda3\\envs\\myenv\\lib\\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",
      "Loaded model from disk\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import os\n",
    "Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_36.json')),'r')\n",
    "loaded_model_json = json_file.read()\n",
    "json_file.close()\n",
    "loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "# load weights into new model\n",
    "loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'models','model_36.h5')))\n",
    "print(\"Loaded model from disk\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\ASUS\\Anaconda3\\envs\\myenv\\lib\\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",
      "Loaded model from disk\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import os\n",
    "Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_44.json')),'r')\n",
    "loaded_model_json = json_file.read()\n",
    "json_file.close()\n",
    "loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "# load weights into new model\n",
    "loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'checkpoints','44','weights.01.hdf5')))\n",
    "print(\"Loaded model from disk\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "ename": "FileNotFoundError",
     "evalue": "[WinError 3] The system cannot find the path specified: 'D:\\\\studies\\\\university\\\\thesis\\\\speech_separation_codes\\\\du16\\\\donesomestuff\\\\test_data\\\\test_log2_norm2'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mFileNotFoundError\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-41-7433f49f4623>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m     15\u001b[0m \u001b[0minput_test\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'test_log2_norm2'\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     16\u001b[0m \u001b[0mimage\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'images'\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 17\u001b[1;33m \u001b[0minputs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlistdir\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnormpath\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mwrite_path\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0minput_test\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     18\u001b[0m \u001b[0minputs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnatsorted\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     19\u001b[0m \u001b[0mmixed\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlistdir\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnormpath\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mwrite_path\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;34m'test_log2'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mFileNotFoundError\u001b[0m: [WinError 3] The system cannot find the path specified: 'D:\\\\studies\\\\university\\\\thesis\\\\speech_separation_codes\\\\du16\\\\donesomestuff\\\\test_data\\\\test_log2_norm2'"
     ]
    }
   ],
   "source": [
    "stoi_eval=[]\n",
    "sdr=[]\n",
    "stoi_mixed=[]\n",
    "sdr_mixed=[]\n",
    "mean_sdr=[]\n",
    "mean_sdr_mixed=[]\n",
    "mean_stoi=[]\n",
    "mean_stoi_mixed=[]\n",
    "# orig_path=os.getcwd()\n",
    "Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "write_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff/test_data'\n",
    "parent = 'results'\n",
    "foldername = 'test'\n",
    "# input_test = 'ftr_refrmd_test'\n",
    "input_test = 'test_log2_norm2'\n",
    "image = 'images'\n",
    "inputs = os.listdir(os.path.normpath(os.path.join(write_path,input_test)))\n",
    "inputs = natsorted(inputs)\n",
    "mixed = os.listdir(os.path.normpath(os.path.join(write_path,'test_log2')))\n",
    "mixed = natsorted(mixed)\n",
    "clean1 = os.listdir(os.path.normpath(os.path.join(write_path,'test_log','clean')))\n",
    "clean1 = natsorted(clean1)\n",
    "clean = clean1.copy()\n",
    "j=0\n",
    "print(len(clean))\n",
    "for i in range(len(inputs)-1):\n",
    "    clean=np.concatenate((clean,clean1),axis=0)\n",
    "phase = os.listdir(os.path.join(write_path,'test_phase2'))\n",
    "phase = natsorted(phase)\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername)):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername,'pesq')):\n",
    "#     print('creating pesq')\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername,'pesq'))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername,'sdr')):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername,'sdr'))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername,'stoi')):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername,'stoi'))\n",
    "for filename1,filename2,filename4 in zip(inputs,phase,mixed):\n",
    "    print(filename1)\n",
    "    print(filename2)\n",
    "    print(filename4)\n",
    "    a = os.listdir(os.path.join(write_path,input_test,filename1))\n",
    "    b = os.listdir(os.path.join(write_path,'test_phase2',filename2))\n",
    "    print('blah')\n",
    "    c = os.listdir(os.path.join(write_path,'test_log2',filename4))\n",
    "    stoi_eval=[]\n",
    "    pesq_eval=[]\n",
    "    sdr=[]\n",
    "    stoi_mixed=[]\n",
    "    sdr_mixed=[]\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,filename1)):\n",
    "        os.mkdir(os.path.join(Data_path,parent,foldername,filename1))\n",
    "    for filename11,filename22,filename33,filename44 in zip(a,b,clean,c):\n",
    "        X_log=np.loadtxt(os.path.join(write_path,input_test,filename1,filename11),delimiter=',')\n",
    "        X_phase=np.loadtxt(os.path.join(write_path,'test_phase2',filename2,filename22),delimiter=',')\n",
    "        target = np.loadtxt(os.path.join(write_path,'test_log','clean',filename33),delimiter=',')\n",
    "        framed_data=librosa.core.stft(target, n_fft=512, hop_length=256, win_length=512, window='hann')\n",
    "        abslt=np.absolute(framed_data)**2\n",
    "        dft_signal=np.log10(abslt+1e-7*np.ones(np.shape(abslt)))\n",
    "        mixed_files=np.loadtxt(os.path.join(write_path,'test_log2',filename4,filename44),delimiter=',')\n",
    "        prediction = loaded_model.predict(X_log)\n",
    "        recon_out = reconstruct(prediction, X_phase)\n",
    "        recon_mixed = reconstruct(mixed_files, X_phase)\n",
    "        recon_clean = reconstruct(dft_signal.T, X_phase)\n",
    "        sf.write(os.path.normpath(os.path.join(Data_path,parent,foldername,filename1)+'\\\\'+filename11.replace('.txt','')+'.wav'),recon_out,16000)\n",
    "        target = target[0:len(recon_out)]\n",
    "\n",
    "        #target = np.reshape(target,(1,len(recon_out)))\n",
    "        #recon_out=np.reshape(recon_out,(1,len(recon_out)))\n",
    "#         pesq_eval.append(get_pesq(target, recon_out,16000))\n",
    "        sdr.append(mir_eval.separation.bss_eval_sources(target, recon_out, compute_permutation=False)[0][0])\n",
    "        stoi_eval.append(stoi(target, recon_out, 16000, extended=False))\n",
    "        sdr_mixed.append(mir_eval.separation.bss_eval_sources(target, recon_mixed, compute_permutation=False)[0][0])\n",
    "        stoi_mixed.append(stoi(target, recon_mixed, 16000, extended=False))\n",
    "    mean_stoi.append(np.mean(stoi_eval))\n",
    "    mean_stoi_mixed.append(np.mean(stoi_mixed))\n",
    "    mean_sdr.append(np.mean(sdr))\n",
    "    mean_sdr_mixed.append(np.mean(sdr_mixed))\n",
    "mean_sdr2 = change_order(mean_sdr)\n",
    "mean_stoi2 = change_order(mean_stoi)\n",
    "mean_sdr_mixed2 = change_order(mean_sdr_mixed)\n",
    "mean_stoi_mixed2 = change_order(mean_stoi_mixed)\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'predictedsdr.txt',mean_sdr2, fmt='%1.4f')\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'mixedsdr.txt',mean_sdr_mixed2, fmt='%1.4f')\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'predictedstoi.txt',mean_stoi2, fmt='%1.4f')\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'mixedstoi.txt',mean_stoi_mixed2, fmt='%1.4f')\n",
    "if not os.path.exists(os.path.join(Data_path, parent, image, foldername)):\n",
    "    os.makedirs(os.path.join(Data_path, parent, image, foldername))\n",
    "plt.plot([-9,-6,-3,0,3,6], mean_sdr2, 'r-o', [-9,-6,-3,0,3,6], mean_sdr_mixed2, 'b-o')\n",
    "plt.gca().legend(('pred','mix'))\n",
    "x=[-9,-6,-3,0,3,6]\n",
    "\n",
    "plt.show\n",
    "plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'sdr.png')))\n",
    "plt.close()\n",
    "plt.plot([-9,-6,-3,0,3,6], mean_stoi2, 'r-o', [-9,-6,-3,0,3,6], mean_stoi_mixed2, 'b-o')\n",
    "plt.gca().legend(('pred','mix'))\n",
    "\n",
    "plt.show\n",
    "plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'stoi.png')))\n",
    "plt.close()\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.6490065656370271,\n",
       " 0.7207940089558972,\n",
       " 0.7866405030576905,\n",
       " 0.5752559945672273,\n",
       " 0.5042776681376199,\n",
       " 0.44009095247383945]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_stoi_mixed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded model from disk\n",
      "44\n",
      "12\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "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": [
    "model_num= 44\n",
    "ckpt_num= '03'\n",
    "import tensorflow as tf\n",
    "import os\n",
    "\n",
    "Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "foldername = 'results_'+str(model_num)+'_2'\n",
    "json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "loaded_model_json = json_file.read()\n",
    "json_file.close()\n",
    "loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "# load weights into new model\n",
    "loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.h5')))\n",
    "print(\"Loaded model from disk\")\n",
    "print(model_num)\n",
    "\n",
    "# Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "# foldername = 'results_'+str(model_num)+'_3'\n",
    "# json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "# loaded_model_json = json_file.read()\n",
    "# json_file.close()\n",
    "# loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "# # load weights into new model\n",
    "# loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.h5')))\n",
    "# print(\"Loaded model from disk\")\n",
    "# print(model_num)\n",
    "\n",
    "# import tensorflow as tf\n",
    "# import os\n",
    "# foldername = 'results_'+str(model_num)+'_ckpt'+ckpt_num\n",
    "# Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "# json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "# loaded_model_json = json_file.read()\n",
    "# json_file.close()\n",
    "# loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "# # load weights into new model\n",
    "# loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'checkpoints',str(model_num),'weights.'+ckpt_num+'.hdf5')))\n",
    "# print(\"Loaded model from disk\")\n",
    "\n",
    "stoi_eval=[]\n",
    "sdr=[]\n",
    "stoi_mixed=[]\n",
    "sdr_mixed=[]\n",
    "mean_sdr=[]\n",
    "mean_sdr_mixed=[]\n",
    "mean_stoi=[]\n",
    "mean_stoi_mixed=[]\n",
    "# orig_path=os.getcwd()\n",
    "Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "write_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff/test_data'\n",
    "parent = 'results'\n",
    "# foldername = 'results_'+str(model_num)\n",
    "# foldername = 'test'\n",
    "foldername2 = 'recon_mixed'\n",
    "# input_test = 'ftr_refrmd_test_w7_norm_withtrain'\n",
    "input_test = 'test_log2_norm_withtrain'\n",
    "image = 'images'\n",
    "inputs = os.listdir(os.path.normpath(os.path.join(write_path,input_test)))\n",
    "inputs = natsorted(inputs)\n",
    "mixed = os.listdir(os.path.normpath(os.path.join(write_path,'test_log2')))\n",
    "mixed = natsorted(mixed)\n",
    "clean1 = os.listdir(os.path.normpath(os.path.join(write_path,'test_log','clean')))\n",
    "clean1 = natsorted(clean1)\n",
    "clean = clean1.copy()\n",
    "j=0\n",
    "print(len(clean))\n",
    "for i in range(len(inputs)-1):\n",
    "    clean=np.concatenate((clean,clean1),axis=0)\n",
    "phase = os.listdir(os.path.join(write_path,'test_phase2'))\n",
    "phase = natsorted(phase)\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername)):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername2)):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername2))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername,'pesq')):\n",
    "#     print('creating pesq')\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername,'pesq'))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername,'sdr')):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername,'sdr'))\n",
    "if not os.path.exists(os.path.join(Data_path,parent,foldername,'stoi')):\n",
    "    os.makedirs(os.path.join(Data_path,parent,foldername,'stoi'))\n",
    "for filename1,filename2,filename4 in zip(inputs,phase,mixed):\n",
    "#     print(filename1)\n",
    "#     print(filename2)\n",
    "#     print(filename4)\n",
    "    a = os.listdir(os.path.join(write_path,input_test,filename1))\n",
    "    b = os.listdir(os.path.join(write_path,'test_phase2',filename2))\n",
    "    c = os.listdir(os.path.join(write_path,'test_log2',filename4))\n",
    "    stoi_eval=[]\n",
    "    pesq_eval=[]\n",
    "    sdr=[]\n",
    "    stoi_mixed=[]\n",
    "    sdr_mixed=[]\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,filename1)):\n",
    "        os.mkdir(os.path.join(Data_path,parent,foldername,filename1))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername2,filename1)):\n",
    "        os.mkdir(os.path.join(Data_path,parent,foldername2,filename1))\n",
    "    for filename11,filename22,filename33,filename44 in zip(a,b,clean,c):\n",
    "        X_log=np.loadtxt(os.path.join(write_path,input_test,filename1,filename11),delimiter=',')\n",
    "        X_phase=np.loadtxt(os.path.join(write_path,'test_phase2',filename2,filename22),delimiter=',')\n",
    "        target = np.loadtxt(os.path.normpath(os.path.join(write_path,'test_log','clean',filename33)),dtype='float32',delimiter=',')\n",
    "        framed_data=librosa.core.stft(target, n_fft=512, hop_length=256, win_length=512, window='hann')\n",
    "        abslt=np.absolute(framed_data)**2\n",
    "        dft_signal=np.log10(abslt+(1e-7*np.ones(np.shape(abslt))))\n",
    "        mixed_files=np.loadtxt(os.path.join(write_path,'test_log2',filename4,filename44),delimiter=',')\n",
    "        prediction = loaded_model.predict(X_log)\n",
    "        recon_out = reconstruct(prediction, X_phase)\n",
    "        recon_mixed = reconstruct(mixed_files, X_phase)\n",
    "        recon_clean = reconstruct(dft_signal.T, X_phase)\n",
    "        sf.write(os.path.normpath(os.path.join(Data_path,parent,foldername,filename1)+'\\\\'+filename11.replace('.txt','')+'.wav'),recon_out,16000)\n",
    "        target = target[0:len(recon_out)]\n",
    "        #target = np.reshape(target,(1,len(recon_out)))\n",
    "        #recon_out=np.reshape(recon_out,(1,len(recon_out)))\n",
    "#         pesq_eval.append(get_pesq(target, recon_out,16000))\n",
    "        sdr.append(mir_eval.separation.bss_eval_sources(target, recon_out, compute_permutation=False)[0][0])\n",
    "        stoi_eval.append(stoi(target, recon_out, 16000, extended=False))\n",
    "        sdr_mixed.append(mir_eval.separation.bss_eval_sources(target, recon_mixed, compute_permutation=False)[0][0])\n",
    "        stoi_mixed.append(stoi(target, recon_mixed, 16000, extended=False))\n",
    "        sf.write(os.path.normpath(os.path.join(Data_path,parent,foldername2,filename1)+'\\\\'+filename11.replace('.txt','')+'.wav'),recon_mixed,16000)\n",
    "        sf.write(os.path.normpath(os.path.join(Data_path,parent,foldername2,'clean')+'\\\\'+filename11.replace('.txt','')+'.wav'),target,16000)\n",
    "    mean_stoi.append(np.mean(stoi_eval))\n",
    "    mean_stoi_mixed.append(np.mean(stoi_mixed))\n",
    "    mean_sdr.append(np.mean(sdr))\n",
    "    mean_sdr_mixed.append(np.mean(sdr_mixed))\n",
    "mean_sdr2 = change_order(mean_sdr)\n",
    "mean_stoi2 = change_order(mean_stoi)\n",
    "mean_sdr_mixed2 = change_order(mean_sdr_mixed)\n",
    "mean_stoi_mixed2 = change_order(mean_stoi_mixed)\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'predictedsdr.txt',mean_sdr2, fmt='%1.4f')\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'mixedsdr.txt',mean_sdr_mixed2, fmt='%1.4f')\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'predictedstoi.txt',mean_stoi2, fmt='%1.4f')\n",
    "np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'mixedstoi.txt',mean_stoi_mixed2, fmt='%1.4f')\n",
    "if not os.path.exists(os.path.join(Data_path, parent, image, foldername)):\n",
    "    os.makedirs(os.path.join(Data_path, parent, image, foldername))\n",
    "plt.plot([-9,-6,-3,0,3,6], mean_sdr2, 'r-o', [-9,-6,-3,0,3,6], mean_sdr_mixed2, 'b-o')\n",
    "plt.gca().legend(('pred','mix'))\n",
    "x=[-9,-6,-3,0,3,6]\n",
    "\n",
    "plt.show()\n",
    "plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'sdr.png')))\n",
    "plt.close()\n",
    "plt.plot([-9,-6,-3,0,3,6], mean_stoi2, 'r-o', [-9,-6,-3,0,3,6], mean_stoi_mixed2, 'b-o')\n",
    "plt.gca().legend(('pred','mix'))\n",
    "\n",
    "plt.show()\n",
    "plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'stoi.png')))\n",
    "plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.46245345, 0.5351304 , 0.61420762, 0.69387352, 0.76800348,\n",
       "       0.83232713])"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_stoi_mixed2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded model from disk\n",
      "45\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "46\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "47\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "48\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "49\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "50\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "51\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n",
      "Loaded model from disk\n",
      "52\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n",
      "3\n",
      "3\n",
      "3\n",
      "blah\n",
      "6\n",
      "6\n",
      "6\n",
      "blah\n",
      "m3\n",
      "m3\n",
      "m3\n",
      "blah\n",
      "m6\n",
      "m6\n",
      "m6\n",
      "blah\n",
      "m9\n",
      "m9\n",
      "m9\n",
      "blah\n"
     ]
    }
   ],
   "source": [
    "for model_num in range(45,53):\n",
    "    ckpt_num= '03'\n",
    "    import tensorflow as tf\n",
    "    import os\n",
    "\n",
    "    Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "    foldername = 'results_'+str(model_num)+'_test'\n",
    "    json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "    loaded_model_json = json_file.read()\n",
    "    json_file.close()\n",
    "    loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "    # load weights into new model\n",
    "    loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.h5')))\n",
    "    print(\"Loaded model from disk\")\n",
    "    print(model_num)\n",
    "\n",
    "    # Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "    # foldername = 'results_'+str(model_num)+'_3'\n",
    "    # json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "    # loaded_model_json = json_file.read()\n",
    "    # json_file.close()\n",
    "    # loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "    # # load weights into new model\n",
    "    # loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.h5')))\n",
    "    # print(\"Loaded model from disk\")\n",
    "    # print(model_num)\n",
    "\n",
    "    # import tensorflow as tf\n",
    "    # import os\n",
    "    # foldername = 'results_'+str(model_num)+'_ckpt'+ckpt_num\n",
    "    # Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "    # json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "    # loaded_model_json = json_file.read()\n",
    "    # json_file.close()\n",
    "    # loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "    # # load weights into new model\n",
    "    # loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'checkpoints',str(model_num),'weights.'+ckpt_num+'.hdf5')))\n",
    "    # print(\"Loaded model from disk\")\n",
    "\n",
    "    stoi_eval=[]\n",
    "    sdr=[]\n",
    "    stoi_mixed=[]\n",
    "    sdr_mixed=[]\n",
    "    mean_sdr=[]\n",
    "    mean_sdr_mixed=[]\n",
    "    mean_stoi=[]\n",
    "    mean_stoi_mixed=[]\n",
    "    # orig_path=os.getcwd()\n",
    "    Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "    write_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff/test_data2'\n",
    "    parent = 'results'\n",
    "    # foldername = 'results_'+str(model_num)\n",
    "    # foldername = 'test'\n",
    "#     input_test = 'ftr_refrmd_test_w7_norm2'\n",
    "    input_test = 'test_log2_norm2'\n",
    "    image = 'images'\n",
    "    inputs = os.listdir(os.path.normpath(os.path.join(write_path,input_test)))\n",
    "    inputs = natsorted(inputs)\n",
    "    mixed = os.listdir(os.path.normpath(os.path.join(write_path,'test_log2')))\n",
    "    mixed = natsorted(mixed)\n",
    "    clean1 = os.listdir(os.path.normpath(os.path.join(write_path,'test_log','clean')))\n",
    "    clean1 = natsorted(clean1)\n",
    "    clean = clean1.copy()\n",
    "    j=0\n",
    "    print(len(clean))\n",
    "    for i in range(len(inputs)-1):\n",
    "        clean=np.concatenate((clean,clean1),axis=0)\n",
    "    phase = os.listdir(os.path.join(write_path,'test_phase2'))\n",
    "    phase = natsorted(phase)\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername)):\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,'pesq')):\n",
    "    #     print('creating pesq')\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername,'pesq'))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,'sdr')):\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername,'sdr'))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,'stoi')):\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername,'stoi'))\n",
    "    for filename1,filename2,filename4 in zip(inputs,phase,mixed):\n",
    "        print(filename1)\n",
    "        print(filename2)\n",
    "        print(filename4)\n",
    "        a = os.listdir(os.path.join(write_path,input_test,filename1))\n",
    "        b = os.listdir(os.path.join(write_path,'test_phase2',filename2))\n",
    "        print('blah')\n",
    "        c = os.listdir(os.path.join(write_path,'test_log2',filename4))\n",
    "        stoi_eval=[]\n",
    "        pesq_eval=[]\n",
    "        sdr=[]\n",
    "        stoi_mixed=[]\n",
    "        sdr_mixed=[]\n",
    "        if not os.path.exists(os.path.join(Data_path,parent,foldername,filename1)):\n",
    "            os.mkdir(os.path.join(Data_path,parent,foldername,filename1))\n",
    "        for filename11,filename22,filename33,filename44 in zip(a,b,clean,c):\n",
    "            X_log=np.loadtxt(os.path.join(write_path,input_test,filename1,filename11),delimiter=',')\n",
    "            X_phase=np.loadtxt(os.path.join(write_path,'test_phase2',filename2,filename22),delimiter=',')\n",
    "            target = np.loadtxt(os.path.normpath(os.path.join(write_path,'test_log','clean',filename33)),dtype='float32',delimiter=',')\n",
    "            framed_data=librosa.core.stft(target, n_fft=512, hop_length=256, win_length=512, window='hann')\n",
    "            abslt=np.absolute(framed_data)**2\n",
    "            dft_signal=np.log10(abslt+(1e-7*np.ones(np.shape(abslt))))\n",
    "            mixed_files=np.loadtxt(os.path.join(write_path,'test_log2',filename4,filename44),delimiter=',')\n",
    "            prediction = loaded_model.predict(X_log)\n",
    "            recon_out = reconstruct(prediction, X_phase)\n",
    "            recon_mixed = reconstruct(mixed_files, X_phase)\n",
    "            recon_clean = reconstruct(dft_signal.T, X_phase)\n",
    "            sf.write(os.path.normpath(os.path.join(Data_path,parent,foldername,filename1)+'\\\\'+filename11.replace('.txt','')+'.wav'),recon_out,16000)\n",
    "            target = target[0:len(recon_out)]\n",
    "            #target = np.reshape(target,(1,len(recon_out)))\n",
    "            #recon_out=np.reshape(recon_out,(1,len(recon_out)))\n",
    "    #         pesq_eval.append(get_pesq(target, recon_out,16000))\n",
    "            sdr.append(mir_eval.separation.bss_eval_sources(target, recon_out, compute_permutation=False)[0][0])\n",
    "            stoi_eval.append(stoi(target, recon_out, 16000, extended=False))\n",
    "            sdr_mixed.append(mir_eval.separation.bss_eval_sources(target, recon_mixed, compute_permutation=False)[0][0])\n",
    "            stoi_mixed.append(stoi(target, recon_mixed, 16000, extended=False))\n",
    "        mean_stoi.append(np.mean(stoi_eval))\n",
    "        mean_stoi_mixed.append(np.mean(stoi_mixed))\n",
    "        mean_sdr.append(np.mean(sdr))\n",
    "        mean_sdr_mixed.append(np.mean(sdr_mixed))\n",
    "    mean_sdr2 = change_order(mean_sdr)\n",
    "    mean_stoi2 = change_order(mean_stoi)\n",
    "    mean_sdr_mixed2 = change_order(mean_sdr_mixed)\n",
    "    mean_stoi_mixed2 = change_order(mean_stoi_mixed)\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'predictedsdr.txt',mean_sdr2, fmt='%1.4f')\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'mixedsdr.txt',mean_sdr_mixed2, fmt='%1.4f')\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'predictedstoi.txt',mean_stoi2, fmt='%1.4f')\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'mixedstoi.txt',mean_stoi_mixed2, fmt='%1.4f')\n",
    "    if not os.path.exists(os.path.join(Data_path, parent, image, foldername)):\n",
    "        os.makedirs(os.path.join(Data_path, parent, image, foldername))\n",
    "    plt.plot([-9,-6,-3,0,3,6], mean_sdr2, 'r-o', [-9,-6,-3,0,3,6], mean_sdr_mixed2, 'b-o')\n",
    "    plt.gca().legend(('pred','mix'))\n",
    "    x=[-9,-6,-3,0,3,6]\n",
    "\n",
    "    plt.show\n",
    "    plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'sdr.png')))\n",
    "    plt.close()\n",
    "    plt.plot([-9,-6,-3,0,3,6], mean_stoi2, 'r-o', [-9,-6,-3,0,3,6], mean_stoi_mixed2, 'b-o')\n",
    "    plt.gca().legend(('pred','mix'))\n",
    "\n",
    "    plt.show\n",
    "    plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'stoi.png')))\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded model from disk\n",
      "6\n",
      "0\n",
      "0\n",
      "0\n",
      "blah\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "Error when checking input: expected dropout_15_input to have shape (257,) but got array with shape (1799,)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-11-5074fd55fc89>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m     75\u001b[0m             \u001b[0mdft_signal\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlog10\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mabslt\u001b[0m\u001b[1;33m+\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m1e-7\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mones\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mabslt\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     76\u001b[0m             \u001b[0mmixed_files\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mloadtxt\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mjoin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mwrite_path\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;34m'test_log2'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mfilename4\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mfilename44\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mdelimiter\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m','\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 77\u001b[1;33m             \u001b[0mprediction\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mloaded_model\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_log\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     78\u001b[0m             \u001b[0mrecon_out\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mreconstruct\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mprediction\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mX_phase\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     79\u001b[0m             \u001b[0mrecon_mixed\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mreconstruct\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmixed_files\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mX_phase\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\myenv\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py\u001b[0m in \u001b[0;36mpredict\u001b[1;34m(self, x, batch_size, verbose, steps, max_queue_size, workers, use_multiprocessing)\u001b[0m\n\u001b[0;32m   1094\u001b[0m       \u001b[1;31m# batch size.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1095\u001b[0m       x, _, _ = self._standardize_user_data(\n\u001b[1;32m-> 1096\u001b[1;33m           x, check_steps=True, steps_name='steps', steps=steps)\n\u001b[0m\u001b[0;32m   1097\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1098\u001b[0m     if (self.run_eagerly or (isinstance(x, iterator_ops.EagerIterator) and\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\myenv\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py\u001b[0m in \u001b[0;36m_standardize_user_data\u001b[1;34m(self, x, y, sample_weight, class_weight, batch_size, check_steps, steps_name, steps, validation_split, shuffle)\u001b[0m\n\u001b[0;32m   2380\u001b[0m         \u001b[0mfeed_input_shapes\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2381\u001b[0m         \u001b[0mcheck_batch_axis\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m  \u001b[1;31m# Don't enforce the batch size.\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2382\u001b[1;33m         exception_prefix='input')\n\u001b[0m\u001b[0;32m   2383\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2384\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0my\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\myenv\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training_utils.py\u001b[0m in \u001b[0;36mstandardize_input_data\u001b[1;34m(data, names, shapes, check_batch_axis, exception_prefix)\u001b[0m\n\u001b[0;32m    360\u001b[0m                 \u001b[1;34m'Error when checking '\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mexception_prefix\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;34m': expected '\u001b[0m \u001b[1;33m+\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    361\u001b[0m                 \u001b[0mnames\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mi\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;34m' to have shape '\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mstr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mshape\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m+\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 362\u001b[1;33m                 ' but got array with shape ' + str(data_shape))\n\u001b[0m\u001b[0;32m    363\u001b[0m   \u001b[1;32mreturn\u001b[0m \u001b[0mdata\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    364\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mValueError\u001b[0m: Error when checking input: expected dropout_15_input to have shape (257,) but got array with shape (1799,)"
     ]
    }
   ],
   "source": [
    "model_num= 45\n",
    "for ckpt_num in range(1,50):\n",
    "    import tensorflow as tf\n",
    "    import os\n",
    "    foldername = 'results_'+str(model_num)+'_ckpt'+str(ckpt_num).zfill(2)\n",
    "    Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "    json_file = open(os.path.normpath(os.path.join(Data_path,'models','model_'+str(model_num)+'.json')),'r')\n",
    "    loaded_model_json = json_file.read()\n",
    "    json_file.close()\n",
    "    loaded_model = tf.keras.models.model_from_json(loaded_model_json)\n",
    "    # load weights into new model\n",
    "    loaded_model.load_weights(os.path.normpath(os.path.join(Data_path,'checkpoints',str(model_num),'weights.'+str(ckpt_num).zfill(2)+'.hdf5')))\n",
    "    print(\"Loaded model from disk\")\n",
    "\n",
    "    stoi_eval=[]\n",
    "    sdr=[]\n",
    "    stoi_mixed=[]\n",
    "    sdr_mixed=[]\n",
    "    mean_sdr=[]\n",
    "    mean_sdr_mixed=[]\n",
    "    mean_stoi=[]\n",
    "    mean_stoi_mixed=[]\n",
    "    # orig_path=os.getcwd()\n",
    "    Data_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff'\n",
    "    write_path = 'D:/studies/university/thesis/speech_separation_codes/du16/donesomestuff/test_data2'\n",
    "    parent = 'results'\n",
    "    # foldername = 'results_'+str(model_num)\n",
    "    # foldername = 'test'\n",
    "    # input_test = 'ftr_refrmd_test_norm'\n",
    "    input_test = 'ftr_refrmd_test_w7_norm'\n",
    "    image = 'images'\n",
    "    inputs = os.listdir(os.path.normpath(os.path.join(write_path,input_test)))\n",
    "    inputs = natsorted(inputs)\n",
    "    mixed = os.listdir(os.path.normpath(os.path.join(write_path,'test_log2')))\n",
    "    mixed = natsorted(mixed)\n",
    "    clean1 = os.listdir(os.path.normpath(os.path.join(write_path,'test_log','clean')))\n",
    "    clean1 = natsorted(clean1)\n",
    "    clean = clean1.copy()\n",
    "    j=0\n",
    "    print(len(clean))\n",
    "    for i in range(len(inputs)-1):\n",
    "        clean=np.concatenate((clean,clean1),axis=0)\n",
    "    phase = os.listdir(os.path.join(write_path,'test_phase2'))\n",
    "    phase = natsorted(phase)\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername)):\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,'pesq')):\n",
    "    #     print('creating pesq')\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername,'pesq'))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,'sdr')):\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername,'sdr'))\n",
    "    if not os.path.exists(os.path.join(Data_path,parent,foldername,'stoi')):\n",
    "        os.makedirs(os.path.join(Data_path,parent,foldername,'stoi'))\n",
    "    for filename1,filename2,filename4 in zip(inputs,phase,mixed):\n",
    "        print(filename1)\n",
    "        print(filename2)\n",
    "        print(filename4)\n",
    "        a = os.listdir(os.path.join(write_path,input_test,filename1))\n",
    "        b = os.listdir(os.path.join(write_path,'test_phase2',filename2))\n",
    "        print('blah')\n",
    "        c = os.listdir(os.path.join(write_path,'test_log2',filename4))\n",
    "        stoi_eval=[]\n",
    "        pesq_eval=[]\n",
    "        sdr=[]\n",
    "        stoi_mixed=[]\n",
    "        sdr_mixed=[]\n",
    "        if not os.path.exists(os.path.join(Data_path,parent,foldername,filename1)):\n",
    "            os.mkdir(os.path.join(Data_path,parent,foldername,filename1))\n",
    "        for filename11,filename22,filename33,filename44 in zip(a,b,clean,c):\n",
    "            X_log=np.loadtxt(os.path.join(write_path,input_test,filename1,filename11),delimiter=',')\n",
    "            X_phase=np.loadtxt(os.path.join(write_path,'test_phase2',filename2,filename22),delimiter=',')\n",
    "            target = np.loadtxt(os.path.normpath(os.path.join(write_path,'test_log','clean',filename33)),dtype='float32',delimiter=',')\n",
    "            framed_data=librosa.core.stft(target, n_fft=512, hop_length=256, win_length=512, window='hann')\n",
    "            abslt=np.absolute(framed_data)**2\n",
    "            dft_signal=np.log10(abslt+(1e-7*np.ones(np.shape(abslt))))\n",
    "            mixed_files=np.loadtxt(os.path.join(write_path,'test_log2',filename4,filename44),delimiter=',')\n",
    "            prediction = loaded_model.predict(X_log)\n",
    "            recon_out = reconstruct(prediction, X_phase)\n",
    "            recon_mixed = reconstruct(mixed_files, X_phase)\n",
    "            recon_clean = reconstruct(dft_signal.T, X_phase)\n",
    "#             sf.write(os.path.normpath(os.path.join(Data_path,parent,foldername,filename1)+'\\\\'+filename11.replace('.txt','')+'.wav'),recon_out,16000)\n",
    "            target = target[0:len(recon_out)]\n",
    "            #target = np.reshape(target,(1,len(recon_out)))\n",
    "            #recon_out=np.reshape(recon_out,(1,len(recon_out)))\n",
    "    #         pesq_eval.append(get_pesq(target, recon_out,16000))\n",
    "            sdr.append(mir_eval.separation.bss_eval_sources(target, recon_out, compute_permutation=False)[0][0])\n",
    "            stoi_eval.append(stoi(target, recon_out, 16000, extended=False))\n",
    "            sdr_mixed.append(mir_eval.separation.bss_eval_sources(target, recon_mixed, compute_permutation=False)[0][0])\n",
    "            stoi_mixed.append(stoi(target, recon_mixed, 16000, extended=False))\n",
    "        mean_stoi.append(np.mean(stoi_eval))\n",
    "        mean_stoi_mixed.append(np.mean(stoi_mixed))\n",
    "        mean_sdr.append(np.mean(sdr))\n",
    "        mean_sdr_mixed.append(np.mean(sdr_mixed))\n",
    "    mean_sdr2 = change_order(mean_sdr)\n",
    "    mean_stoi2 = change_order(mean_stoi)\n",
    "    mean_sdr_mixed2 = change_order(mean_sdr_mixed)\n",
    "    mean_stoi_mixed2 = change_order(mean_stoi_mixed)\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'predictedsdr.txt',mean_sdr2, fmt='%1.4f')\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'sdr')+'\\\\'+'mixedsdr.txt',mean_sdr_mixed2, fmt='%1.4f')\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'predictedstoi.txt',mean_stoi2, fmt='%1.4f')\n",
    "    np.savetxt(os.path.join(Data_path,parent,foldername,'stoi')+'\\\\'+'mixedstoi.txt',mean_stoi_mixed2, fmt='%1.4f')\n",
    "    if not os.path.exists(os.path.join(Data_path, parent, image, foldername)):\n",
    "        os.makedirs(os.path.join(Data_path, parent, image, foldername))\n",
    "    plt.plot([-9,-6,-3,0,3,6], mean_sdr2, 'r-o', [-9,-6,-3,0,3,6], mean_sdr_mixed2, 'b-o')\n",
    "    plt.gca().legend(('pred','mix'))\n",
    "    x=[-9,-6,-3,0,3,6]\n",
    "\n",
    "    plt.show\n",
    "    plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'sdr.png')))\n",
    "    plt.close()\n",
    "    plt.plot([-9,-6,-3,0,3,6], mean_stoi2, 'r-o', [-9,-6,-3,0,3,6], mean_stoi_mixed2, 'b-o')\n",
    "    plt.gca().legend(('pred','mix'))\n",
    "\n",
    "    plt.show\n",
    "    plt.savefig(os.path.normpath(os.path.join(Data_path, parent, image, foldername, 'stoi.png')))\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_pesq(reference, degraded, sample_rate=None, program='pesq'):\n",
    "    \"\"\" Return PESQ quality estimation (two values: PESQ MOS and MOS LQO) based\n",
    "    on reference and degraded speech samples comparison.\n",
    "    Sample rate must be 8000 or 16000 (or can be defined reading reference file\n",
    "    header).\n",
    "    PESQ utility must be installed.\n",
    "    \"\"\"\n",
    "    if not os.path.isfile(reference) or not os.path.isfile(degraded):\n",
    "        raise ValueError('reference or degraded file does not exist')\n",
    "    if not sample_rate:\n",
    "        import wave\n",
    "        w = wave.open(reference, 'r')\n",
    "        sample_rate = w.getframerate()\n",
    "        w.close()\n",
    "    if sample_rate not in (8000, 16000):\n",
    "        raise ValueError('sample rate must be 8000 or 16000')\n",
    "    import subprocess\n",
    "    args = [ program, '+%d' % sample_rate, reference, degraded  ]\n",
    "    pipe = subprocess.Popen(args, stdout=subprocess.PIPE)\n",
    "    out, _ = pipe.communicate()\n",
    "    last_line = out.split('\\n')[-2]\n",
    "    if not last_line.startswith('P.862 Prediction'):\n",
    "        raise ValueError(last_line)\n",
    "    return tuple(map(float, last_line.split()[-2:]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sounddevice as sd\n",
    "sd.play(target,16000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "sd.play(recon_out,16000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "sd.play(recon_mixed,16000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "json_file = open('model.json', 'r')\n",
    "loaded_model_json = json_file.read()\n",
    "json_file.close()\n",
    "loaded_model = model_from_json(loaded_model_json)\n",
    "# load weights into new model\n",
    "loaded_model.load_weights(\"model.h5\")\n",
    "print(\"Loaded model from disk\")\n",
    "loaded_model.get_weights()[7].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'1.13.1'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (myenv)",
   "language": "python",
   "name": "myenv"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}