-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
151 lines (126 loc) · 5.33 KB
/
main.py
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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
#passos para realizar os experimentos
# ler as pastas PositiveRandom e NegativeRandom
# para cada imagem lida, se a imagem existir dentro de dataset, ler o arquivo correspondente
# obter as features do arquivo correspondente usando fft (freq_max, freq_min, duracao, pot_max)
# guardar essas features em um dicionario referenciado pelo nome da pasta
# usar as features para treinar um dfa
# usar as imagens lidas para treinar uma rede convolucional
# execucao : python main.py PositiveRandom
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.multiclass import OneVsRestClassifier
from sklearn.metrics import accuracy_score, f1_score, recall_score, precision_score
from sklearn.model_selection import StratifiedKFold
from sklearn.grid_search import RandomizedSearchCV
from sklearn.neighbors import KNeighborsClassifier
from sklearn.multiclass import OneVsRestClassifier
from sklearn.svm import LinearSVC
from pyAudioAnalysis import audioBasicIO,audioFeatureExtraction
from random import shuffle
import numpy as np
import re
import csv
import os
import sys
import math
import warnings #TODO corrigir future warning
import wave
import scipy.io.wavfile as wavfile
import warnings
warnings.filterwarnings('ignore')
from glob import glob
from os.path import basename, join, exists
mfccRanges = [2,21]
def inductorKNN(x,y):
clf = KNeighborsClassifier()
param_dist = {'n_neighbors': range(1,100),'weights' : ['distance','uniform'], 'metric': ['minkowski','manhattan','euclidean','chebyshev']}
random_search = RandomizedSearchCV(clf,param_distributions=param_dist,n_iter=30,cv=3)
random_search.fit(x,y)
print random_search.best_params_
return random_search.best_estimator_
def featureExtractor(fileName):
[Fs, x] = audioBasicIO.readAudioFile(fileName)
Features = audioFeatureExtraction.stFeatureExtraction(x, Fs,0.001 * Fs, 0.0003 * Fs)
MFCCs = []
for index in range(len(Features)):
MFCCs.append(float(np.mean(Features[index])))
return MFCCs
baseDataset = "/Users/marcostexeira/Downloads/codigoMorcegos/dataset/"
positivePath = sys.argv[1]
audioPaths=[]
imgFiles=[]
imgFilesPerClass=[]
indexDict={}
classesNames={}
audioFileNames={}
classesIds = {}
count=0
for idx, img in enumerate(os.listdir(positivePath)):
if img.find('.png') > 0:
#splitting name
validImg = img[1:]
if 'id' not in validImg:
if validImg.split('_')[0] == 'Carollia':
classesName = validImg.split('_')[0] + ' '+ validImg.split('_')[1]
audioName = validImg.split('_')[2].split('M')[0] + ' M'+ validImg.split('I')[0].split('M')[-1]
imgName = 'c'+ validImg.split('_')[2].split('M')[0] + ' M'+ validImg.split('I')[0].split('M')[-1]
elif not validImg.split('_')[0] == 'Stenodermatinae':
classesName = validImg.split('_')[0] + ' '+ validImg.split('_')[1]
audioName = validImg.split('_')[2] + ' '+ validImg.split('I')[0].split('_')[-1]
if len(validImg.split('_'))>3:
imgName = 'c'+ validImg.split('_')[2] + ' '+ validImg.split('_')[3]
else:
classesName = validImg.split('_')[0]
audioName = validImg.split('_')[1] + ' '+ validImg.split('I')[0].split('_')[-1]
imgName = 'c'+ validImg.split('_')[1] + ' '+ validImg.split('_')[2]
if not classesName in classesNames.keys() and os.path.isfile(baseDataset + classesName+'/'+audioName+'/Spec/Crop/'+imgName):
classesNames[classesName] = count
count+=1
if os.path.isfile(baseDataset + classesName+'/'+audioName+'/Spec/Crop/'+imgName):
if not baseDataset + classesName+'/'+audioName+'/' in audioFileNames.keys():
audioFileNames[baseDataset + classesName+'/'+audioName+'/'] = imgName[1:].replace('png','WAV')
else:
audioFileNames[baseDataset + classesName+'/'+audioName+'/'] = audioFileNames[baseDataset + classesName+'/'+audioName+'/'] + ', ' +imgName[1:].replace('png','WAV')
featuresExtracted={}
classId = {}
X={}
for value in audioFileNames:
arquivos = audioFileNames[value].split(', ')
for arquivo in arquivos:
#signal_enconter(value + arquivo, 55000, 120000, 200)
#featuresExtracted[value+arquivo] = featureExtractor(value + arquivo)
if(classesNames[value.split('/')[6]] in X.keys()):
X[classesNames[value.split('/')[6]]].append(featureExtractor(value + arquivo))
else:
singleList = []
singleList.append(featureExtractor(value + arquivo))
X[classesNames[value.split('/')[6]]] = singleList
# training a classifier
count=0
classifiers = []
acc_vals = []
X_=[]
y_multiclass= []
bestClassifiers = []
for i in range(len(classesNames)):
if len(X[i]) > 8:
for val in X[i]:
X_.append(val)
y_multiclass.append(i)
X_ = np.array(X_)
y_multiclass = np.array(y_multiclass)
skf = StratifiedKFold(n_splits=10, shuffle=True)
for train_index, test_index in skf.split(X_, y_multiclass):
shuffle(test_index)
shuffle(train_index)
X_train, X_test = X_[train_index], X_[test_index]
y_train, y_test = y_multiclass[train_index], y_multiclass[test_index]
#clf = inductorKNN(X_train, y_train)
clf = OneVsRestClassifier(LinearDiscriminantAnalysis()).fit(X_train, y_train)
pred = clf.predict(X_test)
acc_vals.append(accuracy_score(y_test, pred ))
print("accuracy : ", accuracy_score(y_test, pred ) )
print("precision : ", precision_score(y_test, pred, average='weighted' ) )
print("recall : ", recall_score(y_test, pred,average='weighted' ) )
print("f1 : ", f1_score(y_test, pred,average='weighted' ) )
print("\n")
print(np.sum(acc_vals)/10)