HoneyTian commited on
Commit
4f3f6ce
·
1 Parent(s): 897c5df
Files changed (1) hide show
  1. main.py +14 -5
main.py CHANGED
@@ -32,11 +32,17 @@ def get_args():
32
  return args
33
 
34
 
35
- def process_uploaded_file(vad_engine: str, filename: str, silence_time: float = 0.3, longest_activate: float = 3.0) -> Tuple[str, str]:
 
 
 
 
 
 
36
  if vad_engine == "nx_vad":
37
  return run_nx_vad(filename, silence_time, longest_activate)
38
  elif vad_engine == "silero_vad":
39
- return run_silero_vad(filename, silence_time, longest_activate)
40
  else:
41
  return f"vad engine invalid: {vad_engine}", ""
42
 
@@ -61,13 +67,13 @@ def run_nx_vad(filename: str, silence_time: float = 0.3, longest_activate: float
61
  return vad_timestamps, raw_vad_result
62
 
63
 
64
- def run_silero_vad(filename: str, silence_time: float = 0.3, longest_activate: float = 3.0) -> Tuple[str, str]:
65
  filename = Path(filename).as_posix()
66
 
67
  main_logger.info("do silero vad: {}".format(filename))
68
 
69
- cmd = "vad_bins/silero_vad --filename {} --silence_time {}".format(
70
- filename, silence_time
71
  )
72
  raw_vad_result = Command.popen(cmd)
73
 
@@ -119,6 +125,7 @@ def main():
119
  uploaded_vad_engine = gr.Dropdown(choices=["nx_vad", "silero_vad"], value="nx_vad", label="vad_engine")
120
  uploaded_silence_time = gr.Slider(minimum=0.0, maximum=1.0, value=0.3, step=0.01, label="silence time")
121
  uploaded_longest_activate = gr.Slider(minimum=0.0, maximum=20.0, value=3.0, step=0.1, label="longest activate")
 
122
  upload_button = gr.Button("Run VAD", variant="primary")
123
 
124
  with gr.Row():
@@ -132,6 +139,7 @@ def main():
132
  uploaded_file,
133
  uploaded_silence_time,
134
  uploaded_longest_activate,
 
135
  ],
136
  outputs=[
137
  uploaded_vad_timestamps,
@@ -147,6 +155,7 @@ def main():
147
  uploaded_file,
148
  uploaded_silence_time,
149
  uploaded_longest_activate,
 
150
  ],
151
  outputs=[
152
  uploaded_vad_timestamps,
 
32
  return args
33
 
34
 
35
+ def process_uploaded_file(
36
+ vad_engine: str,
37
+ filename: str,
38
+ silence_time: float = 0.3,
39
+ longest_activate: float = 3.0,
40
+ uploaded_threshold: float = 0.5,
41
+ ) -> Tuple[str, str]:
42
  if vad_engine == "nx_vad":
43
  return run_nx_vad(filename, silence_time, longest_activate)
44
  elif vad_engine == "silero_vad":
45
+ return run_silero_vad(filename, silence_time, uploaded_threshold)
46
  else:
47
  return f"vad engine invalid: {vad_engine}", ""
48
 
 
67
  return vad_timestamps, raw_vad_result
68
 
69
 
70
+ def run_silero_vad(filename: str, silence_time: float = 0.3, threshold: float = 0.5) -> Tuple[str, str]:
71
  filename = Path(filename).as_posix()
72
 
73
  main_logger.info("do silero vad: {}".format(filename))
74
 
75
+ cmd = "vad_bins/silero_vad --filename {} --silence_time {} --threshold {}".format(
76
+ filename, silence_time, threshold
77
  )
78
  raw_vad_result = Command.popen(cmd)
79
 
 
125
  uploaded_vad_engine = gr.Dropdown(choices=["nx_vad", "silero_vad"], value="nx_vad", label="vad_engine")
126
  uploaded_silence_time = gr.Slider(minimum=0.0, maximum=1.0, value=0.3, step=0.01, label="silence time")
127
  uploaded_longest_activate = gr.Slider(minimum=0.0, maximum=20.0, value=3.0, step=0.1, label="longest activate")
128
+ uploaded_threshold = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, step=0.1, label="threshold")
129
  upload_button = gr.Button("Run VAD", variant="primary")
130
 
131
  with gr.Row():
 
139
  uploaded_file,
140
  uploaded_silence_time,
141
  uploaded_longest_activate,
142
+ uploaded_threshold,
143
  ],
144
  outputs=[
145
  uploaded_vad_timestamps,
 
155
  uploaded_file,
156
  uploaded_silence_time,
157
  uploaded_longest_activate,
158
+ uploaded_threshold,
159
  ],
160
  outputs=[
161
  uploaded_vad_timestamps,