Spaces:
Running
Running
alibabasglab
commited on
Commit
•
6e44725
1
Parent(s):
b66c175
Update app.py
Browse files
app.py
CHANGED
@@ -9,33 +9,8 @@ BATCH_SIZE = 1
|
|
9 |
|
10 |
device = 0 if torch.cuda.is_available() else "cpu"
|
11 |
|
12 |
-
|
13 |
-
|
14 |
-
'PESQ','DNSMOS'
|
15 |
-
])
|
16 |
-
"""
|
17 |
-
# Copied from https://github.com/openai/whisper/blob/c09a7ae299c4c34c5839a76380ae407e7d785914/whisper/utils.py#L50
|
18 |
-
def format_timestamp(seconds: float, always_include_hours: bool = False, decimal_marker: str = "."):
|
19 |
-
if seconds is not None:
|
20 |
-
milliseconds = round(seconds * 1000.0)
|
21 |
-
|
22 |
-
hours = milliseconds // 3_600_000
|
23 |
-
milliseconds -= hours * 3_600_000
|
24 |
-
|
25 |
-
minutes = milliseconds // 60_000
|
26 |
-
milliseconds -= minutes * 60_000
|
27 |
-
|
28 |
-
seconds = milliseconds // 1_000
|
29 |
-
milliseconds -= seconds * 1_000
|
30 |
-
|
31 |
-
hours_marker = f"{hours:02d}:" if always_include_hours or hours > 0 else ""
|
32 |
-
return f"{hours_marker}{minutes:02d}:{seconds:02d}{decimal_marker}{milliseconds:03d}"
|
33 |
-
else:
|
34 |
-
# we have a malformed timestamp so just return it as is
|
35 |
-
return seconds
|
36 |
-
|
37 |
-
|
38 |
-
def score(test_file, ref_file, score_list, return_timestamps):
|
39 |
mySpeechScore = SpeechScore(score_list)
|
40 |
scores = mySpeechScore(test_path=test_file, reference_path=ref_file, window=None, score_rate=16000, return_mean=False)
|
41 |
return scores
|
@@ -48,25 +23,50 @@ file_score = gr.Interface(
|
|
48 |
inputs=[
|
49 |
gr.Audio(sources=["upload"], label="test file", type="filepath"),
|
50 |
gr.Audio(sources=["upload"], label="reference file", type="filepath"),
|
51 |
-
#gr.Radio(["without reference", "with reference"], label="Task", info="choose non-instrusive or instrusive scoring"),
|
52 |
-
#gr.Checkbox(default=False, label="DNSMOS"),
|
53 |
-
#gr.Checkbox(default=False, label="PESQ"),
|
54 |
gr.Dropdown(
|
55 |
-
["DNSMOS", "
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
),
|
57 |
],
|
58 |
outputs="text",
|
59 |
-
|
60 |
-
#theme="huggingface",
|
61 |
-
title="Score speech from a file",
|
62 |
description=(
|
63 |
-
"Score
|
64 |
-
" commonly used speech quality
|
65 |
" of arbitrary length."
|
66 |
),
|
67 |
)
|
68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
with demo:
|
70 |
-
gr.TabbedInterface([file_score], ["Score
|
71 |
|
72 |
demo.launch()
|
|
|
9 |
|
10 |
device = 0 if torch.cuda.is_available() else "cpu"
|
11 |
|
12 |
+
def score(test_file, ref_file, score_list_nis, score_list_is, return_timestamps):
|
13 |
+
score_list = score_list_nis + score_list_is
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
mySpeechScore = SpeechScore(score_list)
|
15 |
scores = mySpeechScore(test_path=test_file, reference_path=ref_file, window=None, score_rate=16000, return_mean=False)
|
16 |
return scores
|
|
|
23 |
inputs=[
|
24 |
gr.Audio(sources=["upload"], label="test file", type="filepath"),
|
25 |
gr.Audio(sources=["upload"], label="reference file", type="filepath"),
|
|
|
|
|
|
|
26 |
gr.Dropdown(
|
27 |
+
["DNSMOS", "SRMR"], value=["DNSMOS", "SRMR"], multiselect=True, label="Non-Intrusive Scores", info="Choose scores to include, no reference audio is required."
|
28 |
+
),
|
29 |
+
gr.Dropdown(
|
30 |
+
["PESQ", 'NB_PESQ', 'STOI', 'SISDR',
|
31 |
+
'FWSEGSNR', 'LSD', 'BSSEval', 'DNSMOS',
|
32 |
+
'SNR', 'SSNR', 'LLR', 'CSIG', 'CBAK',
|
33 |
+
'COVL', 'MCD'], value=["PESQ", "STOI"], multiselect=True, label="Intrusive Scores", info="Choose scores to include, reference audio is required."
|
34 |
),
|
35 |
],
|
36 |
outputs="text",
|
37 |
+
title="Score speech quality for an audio clip",
|
|
|
|
|
38 |
description=(
|
39 |
+
"Score speech quality with the click of a button! Demo includes the"
|
40 |
+
" commonly used speech quality assessments for the audio file"
|
41 |
" of arbitrary length."
|
42 |
),
|
43 |
)
|
44 |
|
45 |
+
mic_score = gr.Interface(
|
46 |
+
fn=score,
|
47 |
+
inputs=[
|
48 |
+
gr.Audio(sources=["microphone"],
|
49 |
+
waveform_options=gr.WaveformOptions(
|
50 |
+
waveform_color="#01C6FF",
|
51 |
+
waveform_progress_color="#0066B4",
|
52 |
+
skip_length=2,
|
53 |
+
show_controls=False,
|
54 |
+
),
|
55 |
+
),
|
56 |
+
gr.Dropdown(
|
57 |
+
["DNSMOS", "SRMR"], value=["DNSMOS", "SRMR"], multiselect=True, label="Non-Intrusive Scores", info="Choose scores to include, no reference audio is required."
|
58 |
+
),
|
59 |
+
],
|
60 |
+
outputs="text",
|
61 |
+
title="Test microphone quality using speech score",
|
62 |
+
description=(
|
63 |
+
"Score your microphone quality with the click of a button!"
|
64 |
+
" Uses the most popular method to test your microphone quality"
|
65 |
+
" with a short speech clip."
|
66 |
+
),
|
67 |
+
)
|
68 |
+
|
69 |
with demo:
|
70 |
+
gr.TabbedInterface([mic_score, file_score], ["Score Microphone Quality", "Score Speech Quality"])
|
71 |
|
72 |
demo.launch()
|