SEALI/autocomposer_LSTM_mfccs_mean_var.py at master · Atsintli/SEALI · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
import numpy as np
import matplotlib.pyplot as plt
import soundcard as sc
from struct import unpack
from essentia.streaming import *
from essentia import Pool, run, array, reset
from scipy.special import softmax
from essentia import INFO
from feature_extract import *
from functools import reduce
from toolz import assoc
import argparse
import math
import requests # importing the requests library
from pythonosc import dispatcher
from pythonosc import osc_server
from pythonosc import udp_client
import json
from utils import *
import soundfile
from scipy import spatial

input_data = extract_all_mfccs(sorted(glob.glob('segments_short/' + "*.wav")[0:10]))

def getProps(props, dict):
    return map(lambda prop: dict[prop], props)

def concat_features(input_data):
    features = list(map(lambda data:
               list(tz.concat(getProps(
                   ['mfccMean', 'mfccVar'],
                   data))),
    input_data))
    return features


def tf_handler(mfccs):
  headers = {"content-type": "application/json"}
  data = {"instances": [mfccs]}
  #print ("input:",mfccs)
  r = requests.post(url = "http://localhost:8501/v1/models/improv_class:predict", data=json.dumps(data), headers=headers)
  response = r.json()
  #print(response)
  data = response["predictions"]
  print(data)
  return data[0]

i=0
mfccs = concat_features(input_data)
mfcc_results = mfccs

while i < 5:
    data_result = mfcc_results[-9:]
    print("new_MFCCS", data_result)
    mfcc_results.append(tf_handler(mfcc_results[-9:]))
    i = i+1

with open('anotated_mfcc_mean_var.json') as f:
  jsonData = json.load(f)

myData = jsonData.values()# [[{mfccMean: [], className: "1", fileName: ''}], [{mfccMean: [], className: "2", fileName: ''}]]
flatten = lambda t: [item for sublist in t for item in sublist]
myFlatData = flatten(myData) # [{mfccMean: [], className: "1", fileName: ''}, {mfccMean: [], className: "2", fileName: ''}]
#print("soy flaten", myFlatData)

def reducer( mfcc, acc, fileData):
  diff = np.linalg.norm(np.array(fileData['mfccMean']+fileData['mfccVar']) - np.array(mfcc))
  #print (fileData['mfccMean']+fileData['mfccVar'])
  #diff = spatial.distance_matrix(np.array(fileData['mfccMean']), np.array(mfcc))
  if acc==None:
    return tz.assoc(fileData, 'diff', diff)
  else:
    if acc['diff'] <= diff:
        return acc
    else:
        return tz.assoc(fileData, 'diff', diff)

def getClosestCandidate(mfcc):
  return reduce(lambda acc, fileData: reducer(mfcc, acc, fileData), myFlatData, None)

result = list(map(getClosestCandidate,mfcc_results))[10:]

#print (list(map(lambda x: x['file'], result)))
#print(len(result))

def dedupe(tracklist):
    acc = []
    for el in tracklist:
        if len(acc) == 0:
            acc.append(el)
            continue;
        if acc[-1]["file"] != el["file"]:
            acc.append(el)
    return acc

def concatenateFiles(fileName):
    folder = 'Autocomposer/'
    if not os.path.exists(folder):
        os.makedirs(folder)
    files = []

    for i in range(1):
        audiototal = np.array([])
        for elements in fileName:
            num = ('segments_short/' + elements)
            for audio_files in sorted(glob.glob(num + "*.wav")):
                print("Escribiendo " + audio_files)
                y, sr = librosa.load(audio_files)
                audiototal = np.append(audiototal, y)
                soundfile.write(folder + "test_mfccs_mean_var" + ".wav", audiototal, sr)

lastResult = dedupe(result)
lista = list(map(lambda x: x['file'], lastResult))
concatenateFiles(lista)