Audio Tagging

The goal of audio tagging is to assign tags (one or many) to audio signal. Tags can be considered as sound events without timing information (onset or offset).

Metrics

Main functions:

Function sed_eval.audio_tag.AudioTaggingMetrics.evaluate takes as a parameter tag lists, (use dcase_util.containers.MetaDataContainer to read them from a file), and probability lists (use dcase_util.containers.ProbabilityContainer to read them from a file).

Usage example:

  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
208
209
210
211
212
import sed_eval
import dcase_util

reference_tag_list = dcase_util.containers.MetaDataContainer([
    {
        'filename': 'test1.wav',
        'tags': 'cat,dog'
    },
    {
        'filename': 'test2.wav',
        'tags': 'dog'
    },
    {
        'filename': 'test3.wav',
        'tags': 'bird,cat'
    },
    {
        'filename': 'test4.wav',
        'tags': 'cat'
    },
    {
        'filename': 'test5.wav',
        'tags': 'bird,speech'
    },
    {
        'filename': 'test6.wav',
        'tags': 'dog,speech'
    },
    {
        'filename': 'test7.wav',
        'tags': 'speech'
    },
])

estimated_tag_probabilities = dcase_util.containers.ProbabilityContainer([
    {
        'filename': 'test1.wav',
        'label': 'bird',
        'probability': 0.2
    },
    {
        'filename': 'test1.wav',
        'label': 'cat',
        'probability': 0.99
    },
    {
        'filename': 'test1.wav',
        'label': 'dog',
        'probability': 0.88
    },
    {
        'filename': 'test1.wav',
        'label': 'speech',
        'probability': 0.01
    },

    {
        'filename': 'test2.wav',
        'label': 'bird',
        'probability': 0.1
    },
    {
        'filename': 'test2.wav',
        'label': 'cat',
        'probability': 0.3
    },
    {
        'filename': 'test2.wav',
        'label': 'dog',
        'probability': 0.8
    },
    {
        'filename': 'test2.wav',
        'label': 'speech',
        'probability': 0.1
    },


    {
        'filename': 'test3.wav',
        'label': 'bird',
        'probability': 0.7
    },
    {
        'filename': 'test3.wav',
        'label': 'cat',
        'probability': 0.6
    },
    {
        'filename': 'test3.wav',
        'label': 'dog',
        'probability': 0.4
    },
    {
        'filename': 'test3.wav',
        'label': 'speech',
        'probability': 0.3
    },

    {
        'filename': 'test4.wav',
        'label': 'bird',
        'probability': 0.323
    },
    {
        'filename': 'test4.wav',
        'label': 'cat',
        'probability': 0.6
    },
    {
        'filename': 'test4.wav',
        'label': 'dog',
        'probability': 0.56
    },
    {
        'filename': 'test4.wav',
        'label': 'speech',
        'probability': 0.4
    },


    {
        'filename': 'test5.wav',
        'label': 'bird',
        'probability': 0.8
    },
    {
        'filename': 'test5.wav',
        'label': 'cat',
        'probability': 0.7
    },
    {
        'filename': 'test5.wav',
        'label': 'dog',
        'probability': 0.45
    },
    {
        'filename': 'test5.wav',
        'label': 'speech',
        'probability': 0.43
    },


    {
        'filename': 'test6.wav',
        'label': 'bird',
        'probability': 0.9
    },
    {
        'filename': 'test6.wav',
        'label': 'cat',
        'probability': 0.53
    },
    {
        'filename': 'test6.wav',
        'label': 'dog',
        'probability': 0.83
    },
    {
        'filename': 'test6.wav',
        'label': 'speech',
        'probability': 0.95
    },


    {
        'filename': 'test7.wav',
        'label': 'bird',
        'probability': 0.2
    },
    {
        'filename': 'test7.wav',
        'label': 'cat',
        'probability': 0.2
    },
    {
        'filename': 'test7.wav',
        'label': 'dog',
        'probability': 0.89
    },
    {
        'filename': 'test7.wav',
        'label': 'speech',
        'probability': 0.45
    },
])

estimated_tag_list = dcase_util.containers.MetaDataContainer()
for file in estimated_tag_probabilities.unique_files:
    k = estimated_tag_probabilities.filter(filename=file)
    tags = []
    for item in k:
        if item.probability > 0.5:
            tags.append(item.label)

    estimated_tag_list.append(
        {
            'filename': file,
            'tags': tags
        }
    )

tag_evaluator = sed_eval.audio_tag.AudioTaggingMetrics(
    tags=reference_tag_list.unique_tags
)

tag_evaluator.evaluate(
    reference_tag_list=reference_tag_list,
    estimated_tag_list=estimated_tag_list,
    estimated_tag_probabilities=estimated_tag_probabilities
)
print(tag_evaluator)
AudioTaggingMetrics([tags])
AudioTaggingMetrics.evaluate(reference_tag_list) Evaluate estimated against reference
AudioTaggingMetrics.results() All metrics
AudioTaggingMetrics.results_overall_metrics() Overall metrics
AudioTaggingMetrics.results_class_wise_metrics() Class-wise metrics
AudioTaggingMetrics.results_class_wise_average_metrics() Class-wise averaged metrics
AudioTaggingMetrics.result_report_parameters() Report metric parameters
AudioTaggingMetrics.result_report_class_wise() Report class-wise results
AudioTaggingMetrics.result_report_class_wise_average() Report class-wise averages
AudioTaggingMetrics.reset() Reset internal state