-
Notifications
You must be signed in to change notification settings - Fork 1
/
compute_ratios.py
176 lines (157 loc) · 7.95 KB
/
compute_ratios.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
#!/usr/bin/env python
# Copyright Brno University of Technology
# Copyright Gwangju Institute of Science and Technology (Dongkeon Park)
# Licensed under the Apache License, Version 2.0 (the "License")
import numpy as np
from optparse import OptionParser
import os
from rttm_utils import rttm_to_hard_labels
from types import SimpleNamespace
import argparse
def parse_arguments() -> SimpleNamespace:
parser = argparse.ArgumentParser(
description='Compute ratio for silence, speech and overlap from rttms')
parser.add_argument('--in-rttm-dir', type=str, required=True,
help='directory with rttm files')
parser.add_argument('--out-file', type=str, required=True,
help='output file where results are written')
parser.add_argument('--precision', type=float, required=False, default=1000.0,
help='precision used to interpret annotations')
parser.add_argument('--lengths', type=str, required=True,
help='file containing list of lengths per file')
parser.add_argument('--txt-list', type=str, required=True,
help='list of files to process')
parser.add_argument('--uem-file', type=str, required=False,
help='optional uem segments')
args = parser.parse_args()
return args
def main():
args = parse_arguments()
list = [line.rstrip() for line in open(args.txt_list, 'r')]
if not os.path.exists(os.path.dirname(args.out_file)):
os.makedirs(os.path.dirname(args.out_file))
lengths_list = np.loadtxt(args.lengths, dtype=object)
lengths = {}
for line in lengths_list:
name, length = line
lengths[name] = float(length)
if args.uem_file is not None:
uem_list = np.loadtxt(args.uem_file, dtype=object)
uem_info = {}
for line in uem_list:
if line[0] in uem_info.keys():
uem_info[line[0]].append((float(line[2]), float(line[3])))
else:
uem_info[line[0]] = [(float(line[2]), float(line[3]))]
all_sil = 0.0
all_1spk = 0.0
all_2spk = 0.0
all_3spk = 0.0
all_4spk = 0.0
all_seconds = 0.0
all_speech_seconds = 0.0
lengths_sil = []
lengths_1spk = []
lengths_2spk = []
lengths_3spk = []
lengths_4spk = []
with open(args.out_file, 'w') as f:
f.write('Name'.ljust(50)+'Silence (s)'.ljust(18) +
'mean'.ljust(9)+'std'.ljust(13) +
'1 speaker (s)'.ljust(18)+'mean'.ljust(9)+'std'.ljust(13) +
'2 speakers (s)'.ljust(18)+'mean'.ljust(9)+'std'.ljust(13) +
'3 speakers (s)'.ljust(18)+'mean'.ljust(9)+'std'.ljust(13) +
'>3 speakers (s)'.ljust(18)+'mean'.ljust(9)+'std'.ljust(13)+'\n')
for line in list:
key = line
matrix, _ = rttm_to_hard_labels(
args.in_rttm_dir+'/'+key+'.rttm', args.precision, lengths[key])
if args.uem_file is not None:
# Create corresponding mask
mask = np.zeros(matrix.shape[0], dtype=int)
for seg in uem_info[key]:
beg = int(seg[0]*args.precision)
end = int(seg[1]*args.precision)
mask[beg:end] = 1
matrix = matrix[mask == 1,:]
classes = np.sum(matrix, axis=1)
seconds_sil = len(np.where(classes == 0)[0]) / args.precision
seconds_1spk = len(np.where(classes == 1)[0]) / args.precision
seconds_2spk = len(np.where(classes == 2)[0]) / args.precision
seconds_3spk = len(np.where(classes == 3)[0]) / args.precision
seconds_4spk = len(np.where(classes >= 4)[0]) / args.precision
seconds = seconds_sil + seconds_1spk + seconds_2spk + seconds_3spk + seconds_4spk
speech_seconds = seconds - seconds_sil
all_sil += seconds_sil
all_1spk += seconds_1spk
all_2spk += seconds_2spk
all_3spk += seconds_3spk
all_4spk += seconds_4spk
all_seconds += seconds
all_speech_seconds += speech_seconds
changes_positions = np.concatenate(([0], np.where(classes[:-1] != classes[1:])[0]+1, [len(classes)]))
segment_type = classes[changes_positions[:-1]]
segment_length = changes_positions[1:] - changes_positions[:-1]
segments_sil = segment_length[np.where(segment_type == 0)[0]]
segments_1spk = segment_length[np.where(segment_type == 1)[0]]
segments_2spk = segment_length[np.where(segment_type == 2)[0]]
segments_3spk = segment_length[np.where(segment_type == 3)[0]]
segments_4spk = segment_length[np.where(segment_type >= 4)[0]]
mean_sil = np.mean(segments_sil)
std_sil = np.std(segments_sil)
mean_1spk = np.mean(segments_1spk)
std_1spk = np.std(segments_1spk)
mean_2spk = np.mean(segments_2spk)
std_2spk = np.std(segments_2spk)
mean_3spk = np.mean(segments_3spk)
std_3spk = np.std(segments_3spk)
mean_4spk = np.mean(segments_4spk)
std_4spk = np.std(segments_4spk)
lengths_sil.extend(segments_sil)
lengths_1spk.extend(segments_1spk)
lengths_2spk.extend(segments_2spk)
lengths_3spk.extend(segments_3spk)
lengths_4spk.extend(segments_4spk)
f.write(key.ljust(50) + (str(seconds_sil)).ljust(9) +
('('+str(round(100*(seconds_sil/seconds), 2))+'%)').ljust(9) +
(str(round(mean_sil, 2))).ljust(9) +
(str(round(std_sil, 2))).ljust(13) +
(str(seconds_1spk)).ljust(9) +
('('+str(round(100*(seconds_1spk/speech_seconds), 2))+'%)').ljust(9) +
(str(round(mean_1spk, 2))).ljust(9) +
(str(round(std_1spk, 2))).ljust(13) +
(str(seconds_2spk)).ljust(9) +
('('+str(round(100*(seconds_2spk/speech_seconds), 2))+'%)').ljust(9) +
(str(round(mean_2spk, 2))).ljust(9) +
(str(round(std_2spk, 2))).ljust(13) +
(str(seconds_3spk)).ljust(9) +
('('+str(round(100*(seconds_3spk/speech_seconds), 2))+'%)').ljust(9) +
(str(round(mean_3spk, 2))).ljust(9) +
(str(round(std_3spk, 2))).ljust(13) +
(str(seconds_4spk)).ljust(9) +
('('+str(round(100*(seconds_4spk/speech_seconds), 2))+'%)').ljust(9) +
(str(round(mean_4spk, 2))).ljust(9) +
(str(round(std_4spk, 2))).ljust(13)+'\n')
f.write('ALL'.ljust(50) + (str(all_sil)).ljust(9) +
('('+str(round(100*(all_sil/all_seconds), 2))+'%)').ljust(9) +
(str(round(np.mean(lengths_sil), 2))).ljust(9) +
(str(round(np.std(lengths_sil), 2))).ljust(13) +
(str(all_1spk)).ljust(9) +
('('+str(round(100*(all_1spk/all_speech_seconds), 2))+'%)').ljust(9) +
(str(round(np.mean(lengths_1spk), 2))).ljust(9) +
(str(round(np.std(lengths_1spk), 2))).ljust(13) +
(str(all_2spk)).ljust(9) +
('('+str(round(100*(all_2spk/all_speech_seconds), 2))+'%)').ljust(9) +
(str(round(np.mean(lengths_2spk), 2))).ljust(9) +
(str(round(np.std(lengths_2spk), 2))).ljust(13) +
(str(all_3spk)).ljust(9) +
('('+str(round(100*(all_3spk/all_speech_seconds), 2))+'%)').ljust(9) +
(str(round(np.mean(lengths_3spk), 2))).ljust(9) +
(str(round(np.std(lengths_3spk), 2))).ljust(13) +
(str(all_4spk)).ljust(9) +
('('+str(round(100*(all_4spk/all_speech_seconds), 2))+'%)').ljust(9) +
(str(round(np.mean(lengths_4spk), 2))).ljust(9) +
(str(round(np.std(lengths_4spk), 2))).ljust(13) + '\n')
if __name__ == "__main__":
# execute only if run as a script
main()