File size: 3,102 Bytes
c3418e9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Evaluation Configuration# Evaluation Configuration# Evaluation Configuration

# Evaluate model on test sets from raw audio

# Copy this file to eval.yaml and update paths# Evaluate model on test sets from raw audio# Architecture: WavLM + Attentive Pooling + LayerNorm + Deeper Heads



# Model# Copy this file to eval.yaml and update paths# Copy this file to eval.yaml and update paths

model:

  checkpoint: "path/to/best_model"

  name: "microsoft/wavlm-base-plus"

  head_hidden_dim: 256# Model# Model



# Audio Processingmodel:model:

audio:

  sampling_rate: 16000  checkpoint: "path/to/best_model"  checkpoint: "path/to/best_model"

  max_duration: 5

  name: "microsoft/wavlm-base-plus"  name: "microsoft/wavlm-base-plus"

# Evaluation

evaluation:  head_hidden_dim: 256  head_hidden_dim: 256

  batch_size: 32

  dataloader_num_workers: 2



# Data Paths# Audio Processing# Audio Processing

data:

  # === ViSpeech (CSV format) ===audio:audio:

  clean_test_meta: "path/to/metadata/clean_testset.csv"

  clean_test_audio: "path/to/clean_testset"  sampling_rate: 16000  sampling_rate: 16000

  noisy_test_meta: "path/to/metadata/noisy_testset.csv"

  noisy_test_audio: "path/to/noisy_testset"  max_duration: 5  max_duration: 5

  

  # === ViMD (HuggingFace format) ===

  vimd_path: "/path/to/vimd-dataset"

# Evaluation# Evaluation

# Output

output:evaluation:evaluation:

  dir: "output/evaluation"

  save_predictions: true  batch_size: 32  batch_size: 32

  save_confusion_matrix: true

  dataloader_num_workers: 2  dataloader_num_workers: 2

# Label Mappings

labels:

  gender:

    Male: 0# Data Paths# Data Paths (UPDATE THESE PATHS)

    Female: 1

    0: 0data:data:

    1: 1

  dialect:  clean_test_meta: "path/to/metadata/clean_testset.csv"  clean_test_meta: "path/to/metadata/clean_testset.csv"

    North: 0

    Central: 1  clean_test_audio: "path/to/clean_testset"  clean_test_audio: "path/to/clean_testset"

    South: 2

  region_to_dialect:  noisy_test_meta: "path/to/metadata/noisy_testset.csv"  noisy_test_meta: "path/to/metadata/noisy_testset.csv"

    North: 0

    Central: 1  noisy_test_audio: "path/to/noisy_testset"  noisy_test_audio: "path/to/noisy_testset"

    South: 2



# Baseline Comparison (PACLIC 2024 - ResNet34)

baseline:# Output# Output

  gender:

    clean: 98.73output:output:

    noisy: 98.14

  dialect:  dir: "output/evaluation"  dir: "output/evaluation"

    clean: 81.47

    noisy: 74.80  save_predictions: true  save_predictions: true


  save_confusion_matrix: true  save_confusion_matrix: true



# Label Mappings# Label Mappings

labels:labels:

  gender:  gender:

    Male: 0    Male: 0

    Female: 1    Female: 1

    0: 0  dialect:

    1: 1    North: 0

  dialect:    Central: 1

    North: 0    South: 2

    Central: 1

    South: 2# Baseline Comparison (PACLIC 2024 - ResNet34)

baseline:

# Baseline Comparison (PACLIC 2024 - ResNet34)  gender:

baseline:    clean: 98.73

  gender:    noisy: 98.14

    clean: 98.73  dialect:

    noisy: 98.14    clean: 81.47

  dialect:    noisy: 74.80

    clean: 81.47
    noisy: 74.80