-
Notifications
You must be signed in to change notification settings - Fork 4
/
test_for_all.py
207 lines (178 loc) · 9.12 KB
/
test_for_all.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
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
from mir.nn.train import NetworkInterface
import mir.io as io
import datasets
from extractors.cqt import CQTV2,SimpleChordToID
from mir import io,DataEntry
from io_new.chordlab_io import ChordLabIO
from extractors.xhmm_decoder import XHMMDecoder,prob_to_spectrogram
from complex_chord import Chord,ChordTypeLimit,shift_complex_chord_array_list,complex_chord_chop,enum_to_dict,\
TriadTypes,SeventhTypes,NinthTypes,EleventhTypes,ThirteenthTypes
from mir.music_base import NUM_TO_ABS_SCALE
import os
import mir_eval
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
class ExperimentTest():
#ET_confusion_chord = ["Others","N","maj/2","maj/4","13","11","maj9","9","min7","maj7","7","maj6","min6","maj/3","maj/5","min/b3","min/5","maj","min"]
ET_confusion_chord = None #["Others","N","sus2","sus4","hdim7","dim7","min7","minmaj7","maj7","7","maj6","min6","aug","dim","maj","min"]
def __init__(self,isDirectory = False,address_x = "",address_y = ""):
self.address = []
self.address.append(address_x)
self.address.append(address_y)
self.isDirectory = isDirectory
self.correct_durations = [0.0,0.0,0.0,0.0,0.0]
self.wrong_durations = [0.0,0.0,0.0,0.0,0.0]
self.evals = [mir_eval.chord.root,mir_eval.chord.majmin,mir_eval.chord.majmin_inv,mir_eval.chord.sevenths,mir_eval.chord.sevenths_inv]
self.evalStrings = ["Eval for root", "Eval for MajorMinor:","Eval for MajorMinor_inv:","Eval for Sevenths:","Eval for Sevenths_inv:"]
self.confusion_mat = []
self.confusion_total = []
ET_confusion_chord = ExperimentTest.ET_confusion_chord
for i in range(len(ET_confusion_chord)):
c = []
for j in range(len(ET_confusion_chord)):
c.append(0.0)
self.confusion_mat.append(c)
self.confusion_total.append(0.0)
def index_from_chord(self,fchord):
ET_confusion_chord = ExperimentTest.ET_confusion_chord
for i,chord in enumerate(ET_confusion_chord):
if fchord==chord:
return i
return -1
def compare_chord_for_mat(self,x1,x2,duration):
coms_1 = x1.split(":")
coms_2 = x2.split(":")
if len(coms_1) == 1:
coms_1.append(coms_1[0])
if len(coms_2) == 1:
coms_2.append(coms_2[0])
if coms_1[0] == coms_2[0] or coms_1[0] == "N" or coms_1[0] == "X":
idx_1 = self.index_from_chord(coms_1[1])
idx_2 = self.index_from_chord(coms_2[1])
if idx_1 == -1:
idx_1=0
if idx_2 == -1:
# print(x2)
idx_2=0
self.confusion_total[idx_1] = self.confusion_total[idx_1] + duration
self.confusion_mat[idx_1][idx_2] = self.confusion_mat[idx_1][idx_2] + duration
def get_lab(self,file_address):
chordlab = []
with open(file_address,"r") as f:
for lines in f:
lines = lines.strip()
lines = lines.split()
chordlab.append([float(lines[0]),float(lines[1]),lines[2]])
return chordlab
def split_chordlab(self,chordlab):
return (np.array([[data[0],data[1]] for data in chordlab],dtype=np.float64),[data[2] for data in chordlab])
def test(self,gd,est):
name = gd
gd = self.get_lab(gd)
est = self.get_lab(est)
if gd[0][0] < 0:
gd[0][0] = 0
(gd_intervals,gd_labels) = self.split_chordlab(gd)
(est_intervals,est_labels) = self.split_chordlab(est)
est_intervals,est_labels = mir_eval.util.adjust_intervals(est_intervals,est_labels,gd_intervals.min(),gd_intervals.max(),start_label='X',end_label='X')
(intervals,gd_labels,est_labels)=mir_eval.util.merge_labeled_intervals(gd_intervals,gd_labels,est_intervals,est_labels)
durations = mir_eval.util.intervals_to_durations(intervals)
compares = []
scores = []
for evalk in self.evals:
compare = evalk(gd_labels,est_labels)
compares.append(compare)
score = mir_eval.chord.weighted_accuracy(compare,durations)
scores.append(score)
#print(name+":\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f"%(scores[0] * 100.0,scores[1] * 100.0,scores[2] * 100.0,scores[3] * 100.0,scores[4] * 100.0))
for i in range(len(self.correct_durations)):
self.correct_durations[i] += durations[compares[i] > 0].sum()
self.wrong_durations[i] += durations[compares[i] == 0].sum()
for i in range(len(durations)):
self.compare_chord_for_mat(gd_labels[i],est_labels[i],durations[i])
def calc_accuracy(self):
if not 1 == 1:
print("Model File not exist!")
else:
if self.isDirectory:
path = os.listdir(self.address[0])
gds = []
ests = []
for i in path:
gds.append(i)
gds = np.array(gds)
path = os.listdir(self.address[1])
for i in path:
ests.append(i)
ests = np.array(ests)
inters = np.intersect1d(gds,ests)
print("total files: " + str(len(inters)))
for q in inters:
self.test(self.address[0] + q,self.address[1] + q)
else:
self.test(self.address[0],self.address[1])
for i in range(len(self.correct_durations)):
print(self.evalStrings[i],'%.2f%%'%(100.0*self.correct_durations[i]/(self.correct_durations[i]+self.wrong_durations[i])))
def draw_confusion_mat(self):
self.confusion_mat = np.array(self.confusion_mat)
print('Diag ratio unnormalized', self.confusion_mat[1:,1:].diagonal().sum()/self.confusion_mat[1:,1:].sum())
for i,submat in enumerate(self.confusion_mat):
if self.confusion_total[i] == 0:
self.confusion_total[i] = 1
for j,value in enumerate(submat):
self.confusion_mat[i][j] /= self.confusion_total[i]
self.confusion_mat[i][j] = round(self.confusion_mat[i][j],2)
print(self.confusion_mat)
self.confusion_mat = np.array(self.confusion_mat)
print('Diag ratio normalized', self.confusion_mat[1:,1:].diagonal().sum()/self.confusion_mat[1:,1:].sum())
fig,ax = plt.subplots(figsize=(8,6))
im = ax.imshow(self.confusion_mat,cmap = "magma_r",vmin=0.0,vmax=1.0)
cbar = ax.figure.colorbar(im,ax = ax)
cbar.ax.set_ylabel("ratio of confusion",rotation = -90,va = "bottom")
plt.xlabel('estimation')
plt.ylabel('reference')
ax.set_xticks(np.arange(len(ExperimentTest.ET_confusion_chord)))
ax.set_yticks(np.arange(len(ExperimentTest.ET_confusion_chord)))
ax.set_xticklabels(ExperimentTest.ET_confusion_chord)
ax.set_yticklabels(ExperimentTest.ET_confusion_chord)
plt.setp(ax.get_xticklabels(),rotation = 45,ha = "right",rotation_mode = "anchor")
for i in range(len(ExperimentTest.ET_confusion_chord)):
for j in range(len(ExperimentTest.ET_confusion_chord)):
scolor = 'w'
if self.confusion_mat[i][j] < 0.3:
scolor = '#000000'
#value = ax.text(j,i,self.confusion_mat[i][j],ha = "center", va = "center", color = scolor)
#ax.set_title("Confusion Matrix for Chord Recognition")
fig.tight_layout()
from figures import FIG_OUTPUT_PATH
fig.savefig(os.path.join(FIG_OUTPUT_PATH,'confusion_matrix.pdf'), transparent=True, pad_inches=0,bbox_inches='tight')
plt.show()
def extract_quality_list_from_file(filename):
f=open(filename,'r')
lines=[line.strip() for line in f.readlines() if line.strip()!='']
f.close()
result=['Others']
for line in lines:
if(line.startswith('C:')):
result.append(line[2:])
else:
result.append(line)
return result
def main():
from settings import JAM_DATASET_PATH,MY_DATASET_PATH
ExperimentTest.ET_confusion_chord=extract_quality_list_from_file('data/submission_chord_list.txt')
#q = ExperimentTest(True,os.path.join(JAM_DATASET_PATH,'chordlab')+'/',"output/output_joint_chord_net_ismir_flat_v1.0_reweight(1.0,1.0)_s%d.best_hmm_submission/jam/")
q = ExperimentTest(True,os.path.join(JAM_DATASET_PATH,'chordlab')+'/',"output/output_joint_chord_net_ismir_v1.0_triad_only_reweight(1.0,1.0)_s%d.best_hmm_ismir2017/jam/")
#q = ExperimentTest(True,os.path.join(JAM_DATASET_PATH,'chordlab')+'/',"output/output_hook_mirex2019_chordnet_v1.1_fix_s%d.best_hmm_ismir2017/jam/")
q.calc_accuracy()
q.draw_confusion_mat()
def eval_submission(reweight_factor,reweight_max):
from settings import JAM_DATASET_PATH
ExperimentTest.ET_confusion_chord=extract_quality_list_from_file('data/submission_chord_list.txt')
q = ExperimentTest(True,os.path.join(JAM_DATASET_PATH,'chordlab')+'/',"output/output_joint_chord_net_ismir_naive_v1.0_reweight(%.1f,%.1f)"%(reweight_factor,reweight_max)+"_s%d.best_hmm_submission/jam/")
q.calc_accuracy()
q.draw_confusion_mat()
if __name__ == "__main__":
main()
#eval_submission(0.0,10.0)