Update app.py
Browse files
app.py
CHANGED
@@ -2,9 +2,12 @@ import gradio as gr
|
|
2 |
import numpy as np
|
3 |
import torch
|
4 |
import librosa
|
|
|
5 |
from transformers import AutoFeatureExtractor, AutoModelForAudioClassification
|
6 |
import matplotlib.pyplot as plt
|
7 |
from matplotlib.colors import Normalize
|
|
|
|
|
8 |
|
9 |
# Constants
|
10 |
SAMPLING_RATE = 16000
|
@@ -15,26 +18,42 @@ DEFAULT_THRESHOLD = 0.7
|
|
15 |
feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_NAME)
|
16 |
model = AutoModelForAudioClassification.from_pretrained(MODEL_NAME)
|
17 |
|
18 |
-
def
|
19 |
-
"""
|
20 |
-
Process audio and detect anomalies
|
21 |
-
Returns:
|
22 |
-
- classification result
|
23 |
-
- confidence score
|
24 |
-
- spectrogram visualization
|
25 |
-
"""
|
26 |
try:
|
27 |
-
#
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
|
|
35 |
if len(audio.shape) > 1:
|
36 |
-
audio =
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
# Extract features
|
39 |
inputs = feature_extractor(
|
40 |
audio,
|
@@ -50,15 +69,15 @@ def analyze_audio(audio_array, threshold=DEFAULT_THRESHOLD):
|
|
50 |
logits = outputs.logits
|
51 |
probs = torch.softmax(logits, dim=-1)
|
52 |
|
53 |
-
# Get
|
54 |
predicted_class = "Normal" if probs[0][0] > threshold else "Anomaly"
|
55 |
confidence = probs[0][0].item() if predicted_class == "Normal" else 1 - probs[0][0].item()
|
56 |
|
57 |
-
# Create spectrogram
|
58 |
spectrogram = librosa.feature.melspectrogram(
|
59 |
y=audio,
|
60 |
sr=SAMPLING_RATE,
|
61 |
-
n_mels=64,
|
62 |
fmax=8000
|
63 |
)
|
64 |
db_spec = librosa.power_to_db(spectrogram, ref=np.max)
|
@@ -75,18 +94,21 @@ def analyze_audio(audio_array, threshold=DEFAULT_THRESHOLD):
|
|
75 |
fig.colorbar(img, ax=ax, format='%+2.0f dB')
|
76 |
ax.set(title='Mel Spectrogram')
|
77 |
plt.tight_layout()
|
78 |
-
|
|
|
|
|
|
|
79 |
plt.close()
|
80 |
|
81 |
return (
|
82 |
predicted_class,
|
83 |
f"{confidence:.1%}",
|
84 |
-
|
85 |
str(probs.tolist()[0])
|
86 |
)
|
87 |
|
88 |
except Exception as e:
|
89 |
-
return f"Error: {str(e)}", "",
|
90 |
|
91 |
# Gradio interface
|
92 |
with gr.Blocks(title="Industrial Audio Analyzer", theme=gr.themes.Soft()) as demo:
|
@@ -98,16 +120,15 @@ with gr.Blocks(title="Industrial Audio Analyzer", theme=gr.themes.Soft()) as dem
|
|
98 |
with gr.Row():
|
99 |
with gr.Column():
|
100 |
audio_input = gr.Audio(
|
101 |
-
label="Upload Equipment Audio
|
102 |
-
type="
|
103 |
)
|
104 |
threshold = gr.Slider(
|
105 |
minimum=0.5,
|
106 |
maximum=0.95,
|
107 |
step=0.05,
|
108 |
value=DEFAULT_THRESHOLD,
|
109 |
-
label="Anomaly Detection Threshold"
|
110 |
-
info="Higher values reduce false positives but may miss subtle anomalies"
|
111 |
)
|
112 |
analyze_btn = gr.Button("π Analyze Sound", variant="primary")
|
113 |
|
@@ -127,12 +148,10 @@ with gr.Blocks(title="Industrial Audio Analyzer", theme=gr.themes.Soft()) as dem
|
|
127 |
)
|
128 |
|
129 |
gr.Markdown("""
|
130 |
-
|
131 |
-
- Upload audio recordings
|
132 |
-
-
|
133 |
-
-
|
134 |
-
|
135 |
-
**Tip**: For best results, use 5-10 second recordings of steady operation
|
136 |
""")
|
137 |
|
138 |
if __name__ == "__main__":
|
|
|
2 |
import numpy as np
|
3 |
import torch
|
4 |
import librosa
|
5 |
+
import soundfile as sf
|
6 |
from transformers import AutoFeatureExtractor, AutoModelForAudioClassification
|
7 |
import matplotlib.pyplot as plt
|
8 |
from matplotlib.colors import Normalize
|
9 |
+
import tempfile
|
10 |
+
import os
|
11 |
|
12 |
# Constants
|
13 |
SAMPLING_RATE = 16000
|
|
|
18 |
feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_NAME)
|
19 |
model = AutoModelForAudioClassification.from_pretrained(MODEL_NAME)
|
20 |
|
21 |
+
def handle_audio_file(audio_file):
|
22 |
+
"""Handle uploaded audio file and convert to numpy array"""
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
try:
|
24 |
+
# Save to temp file and load with soundfile
|
25 |
+
with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as tmp:
|
26 |
+
tmp.write(audio_file.read())
|
27 |
+
tmp_path = tmp.name
|
28 |
+
|
29 |
+
audio, sr = sf.read(tmp_path)
|
30 |
+
os.unlink(tmp_path) # Clean up temp file
|
31 |
+
|
32 |
+
# Convert to mono if needed
|
33 |
if len(audio.shape) > 1:
|
34 |
+
audio = np.mean(audio, axis=1)
|
35 |
|
36 |
+
return audio, sr
|
37 |
+
except Exception as e:
|
38 |
+
raise ValueError(f"Error processing audio file: {str(e)}")
|
39 |
+
|
40 |
+
def analyze_audio(audio_input, threshold=DEFAULT_THRESHOLD):
|
41 |
+
"""Process audio and detect anomalies"""
|
42 |
+
try:
|
43 |
+
# Handle different input types
|
44 |
+
if isinstance(audio_input, str): # File path
|
45 |
+
audio, sr = handle_audio_file(open(audio_input, 'rb'))
|
46 |
+
elif hasattr(audio_input, 'name'): # Gradio file object
|
47 |
+
audio, sr = handle_audio_file(audio_input)
|
48 |
+
elif isinstance(audio_input, tuple): # Direct numpy array
|
49 |
+
sr, audio = audio_input
|
50 |
+
else:
|
51 |
+
raise ValueError("Unsupported audio input format")
|
52 |
+
|
53 |
+
# Resample if needed
|
54 |
+
if sr != SAMPLING_RATE:
|
55 |
+
audio = librosa.resample(audio, orig_sr=sr, target_sr=SAMPLING_RATE)
|
56 |
+
|
57 |
# Extract features
|
58 |
inputs = feature_extractor(
|
59 |
audio,
|
|
|
69 |
logits = outputs.logits
|
70 |
probs = torch.softmax(logits, dim=-1)
|
71 |
|
72 |
+
# Get results
|
73 |
predicted_class = "Normal" if probs[0][0] > threshold else "Anomaly"
|
74 |
confidence = probs[0][0].item() if predicted_class == "Normal" else 1 - probs[0][0].item()
|
75 |
|
76 |
+
# Create spectrogram
|
77 |
spectrogram = librosa.feature.melspectrogram(
|
78 |
y=audio,
|
79 |
sr=SAMPLING_RATE,
|
80 |
+
n_mels=64,
|
81 |
fmax=8000
|
82 |
)
|
83 |
db_spec = librosa.power_to_db(spectrogram, ref=np.max)
|
|
|
94 |
fig.colorbar(img, ax=ax, format='%+2.0f dB')
|
95 |
ax.set(title='Mel Spectrogram')
|
96 |
plt.tight_layout()
|
97 |
+
|
98 |
+
# Save to temp file
|
99 |
+
spec_path = os.path.join(tempfile.gettempdir(), 'spec.png')
|
100 |
+
plt.savefig(spec_path, bbox_inches='tight')
|
101 |
plt.close()
|
102 |
|
103 |
return (
|
104 |
predicted_class,
|
105 |
f"{confidence:.1%}",
|
106 |
+
spec_path,
|
107 |
str(probs.tolist()[0])
|
108 |
)
|
109 |
|
110 |
except Exception as e:
|
111 |
+
return f"Error: {str(e)}", "", None, ""
|
112 |
|
113 |
# Gradio interface
|
114 |
with gr.Blocks(title="Industrial Audio Analyzer", theme=gr.themes.Soft()) as demo:
|
|
|
120 |
with gr.Row():
|
121 |
with gr.Column():
|
122 |
audio_input = gr.Audio(
|
123 |
+
label="Upload Equipment Audio (.wav)",
|
124 |
+
type="filepath"
|
125 |
)
|
126 |
threshold = gr.Slider(
|
127 |
minimum=0.5,
|
128 |
maximum=0.95,
|
129 |
step=0.05,
|
130 |
value=DEFAULT_THRESHOLD,
|
131 |
+
label="Anomaly Detection Threshold"
|
|
|
132 |
)
|
133 |
analyze_btn = gr.Button("π Analyze Sound", variant="primary")
|
134 |
|
|
|
148 |
)
|
149 |
|
150 |
gr.Markdown("""
|
151 |
+
**Instructions:**
|
152 |
+
- Upload .wav audio recordings (5-10 seconds recommended)
|
153 |
+
- Adjust threshold to control sensitivity
|
154 |
+
- Results show Normal/Anomaly classification with confidence
|
|
|
|
|
155 |
""")
|
156 |
|
157 |
if __name__ == "__main__":
|