Spaces:
Runtime error
Runtime error
Kabatubare
commited on
Commit
•
faee536
1
Parent(s):
1c53cc7
Update app.py
Browse files
app.py
CHANGED
@@ -4,42 +4,26 @@ import torch
|
|
4 |
import numpy as np
|
5 |
import traceback
|
6 |
|
7 |
-
# Function to handle audio data as NumPy arrays
|
8 |
def detect_watermark(audio_data, sample_rate):
|
9 |
try:
|
10 |
-
# Extract the audio array from the tuple (audio_data, sample_rate)
|
11 |
audio_array, _ = audio_data
|
12 |
-
|
13 |
-
# Ensure audio_array is 2D (channels, samples). If it's mono, add an axis.
|
14 |
if audio_array.ndim == 1:
|
15 |
audio_array = np.expand_dims(audio_array, axis=0)
|
16 |
-
|
17 |
-
# Convert NumPy array to tensor
|
18 |
waveform = torch.tensor(audio_array, dtype=torch.float32)
|
19 |
-
|
20 |
-
# Ensure waveform is 2D (batch, channels, samples) for AudioSeal
|
21 |
if waveform.ndim == 2:
|
22 |
waveform = waveform.unsqueeze(0)
|
23 |
-
|
24 |
-
# Initialize and use the AudioSeal detector
|
25 |
detector = AudioSeal.load_detector("audioseal_detector_16bits")
|
26 |
result, message = detector.detect_watermark(waveform, message_threshold=0.5)
|
27 |
-
|
28 |
-
# Interpret and return the detection result
|
29 |
detection_result = "AI-generated" if result else "genuine"
|
30 |
return f"This audio is likely {detection_result} based on watermark detection."
|
31 |
except Exception as e:
|
32 |
-
|
33 |
-
|
34 |
-
full_error_message = f"{error_message}\n{traceback_str}"
|
35 |
-
return full_error_message
|
36 |
|
37 |
interface = gr.Interface(fn=detect_watermark,
|
38 |
inputs=[gr.Audio(label="Upload your audio", type="numpy"),
|
39 |
gr.Number(label="Sample Rate", value=44100, visible=False)],
|
40 |
-
outputs="text"
|
41 |
-
title="Deep Fake Defender: AI Voice Cloning Detection",
|
42 |
-
description="Upload an audio file to check if it's AI-generated or genuine.")
|
43 |
|
44 |
if __name__ == "__main__":
|
45 |
-
interface.launch(
|
|
|
4 |
import numpy as np
|
5 |
import traceback
|
6 |
|
|
|
7 |
def detect_watermark(audio_data, sample_rate):
|
8 |
try:
|
|
|
9 |
audio_array, _ = audio_data
|
|
|
|
|
10 |
if audio_array.ndim == 1:
|
11 |
audio_array = np.expand_dims(audio_array, axis=0)
|
|
|
|
|
12 |
waveform = torch.tensor(audio_array, dtype=torch.float32)
|
|
|
|
|
13 |
if waveform.ndim == 2:
|
14 |
waveform = waveform.unsqueeze(0)
|
|
|
|
|
15 |
detector = AudioSeal.load_detector("audioseal_detector_16bits")
|
16 |
result, message = detector.detect_watermark(waveform, message_threshold=0.5)
|
|
|
|
|
17 |
detection_result = "AI-generated" if result else "genuine"
|
18 |
return f"This audio is likely {detection_result} based on watermark detection."
|
19 |
except Exception as e:
|
20 |
+
error_traceback = traceback.format_exc()
|
21 |
+
return f"Error: {str(e)}\n{error_traceback}"
|
|
|
|
|
22 |
|
23 |
interface = gr.Interface(fn=detect_watermark,
|
24 |
inputs=[gr.Audio(label="Upload your audio", type="numpy"),
|
25 |
gr.Number(label="Sample Rate", value=44100, visible=False)],
|
26 |
+
outputs="text")
|
|
|
|
|
27 |
|
28 |
if __name__ == "__main__":
|
29 |
+
interface.launch()
|