Update app.py
Browse files
app.py
CHANGED
@@ -14,17 +14,6 @@ import io
|
|
14 |
import librosa
|
15 |
import ffmpeg
|
16 |
|
17 |
-
#from torchaudio.io import CodecConfig
|
18 |
-
# import numpy
|
19 |
-
|
20 |
-
# def my_read_file(audio_path, max_second):
|
21 |
-
# signal, sr, audio_length_second = read_as_single_channel_16k(audio_path, default_sr)
|
22 |
-
# if audio_length_second > max_second:
|
23 |
-
# signal = signal[0:default_sr * max_second]
|
24 |
-
# audio_length_second = max_second
|
25 |
-
|
26 |
-
# return signal, sr, audio_length_second
|
27 |
-
|
28 |
def create_default_value():
|
29 |
if "def_value" not in st.session_state:
|
30 |
def_val_npy = np.random.choice([0, 1], size=32 - len_start_bit)
|
@@ -98,18 +87,6 @@ def main():
|
|
98 |
file_extension_ori =".mp3"
|
99 |
file_extension =".wav"
|
100 |
|
101 |
-
#RuntimeError: Could not infer dtype of numpy.float32
|
102 |
-
#wav = torch.tensor(wav3).float() / 32768.0
|
103 |
-
|
104 |
-
#RuntimeError: Numpy is not available
|
105 |
-
# wav = torch.from_numpy(wav3) #/32768.0
|
106 |
-
# wav = wav.unsqueeze(0).unsqueeze(0)
|
107 |
-
# st.markdown("Before unsqueeze mp3")
|
108 |
-
# st.markdown(wav)
|
109 |
-
|
110 |
-
#Unsqueeze for line 176
|
111 |
-
# wav= wav.unsqueeze(0)
|
112 |
-
|
113 |
action = st.selectbox("Select Action", ["Add Watermark", "Detect Watermark"])
|
114 |
|
115 |
if action == "Add Watermark":
|
@@ -205,179 +182,6 @@ def main():
|
|
205 |
|
206 |
except RuntimeError:
|
207 |
st.error("Please input audio with one channel (mono-channel)")
|
208 |
-
|
209 |
-
# if audio_file:
|
210 |
-
# # 保存文件到本地:
|
211 |
-
# # tmp_input_audio_file = os.path.join("/tmp/", audio_file.name)
|
212 |
-
# # st.markdown(tmp_input_audio_file)
|
213 |
-
# # with open(tmp_input_audio_file, "wb") as f:
|
214 |
-
# # f.write(audio_file.getbuffer())
|
215 |
-
# # st.audio(tmp_input_audio_file, format="mp3/wav")
|
216 |
-
|
217 |
-
|
218 |
-
# #1st attempt
|
219 |
-
# #audio_path = " audio_file.name"
|
220 |
-
|
221 |
-
# # audio, sr = torchaudio.load(audio_file)
|
222 |
-
# # st.audio(audio_file, format="audio/mpeg")
|
223 |
-
# # audio= audio.unsqueeze(0)
|
224 |
-
|
225 |
-
# # st.markdown("SR")
|
226 |
-
# # st.markdown(sr)
|
227 |
-
# # st.markdown("after unsqueeze wav or mp3")
|
228 |
-
# # st.markdown(audio)
|
229 |
-
|
230 |
-
# #2nd attempt
|
231 |
-
# # Save file to local storage
|
232 |
-
# tmp_input_audio_file = os.path.join("/tmp/", audio_file.name)
|
233 |
-
# file_extension = os.path.splitext(tmp_input_audio_file)[1].lower()
|
234 |
-
# #st.markdown(file_extension)
|
235 |
-
# if file_extension in [".wav", ".flac"]:
|
236 |
-
# with open("test.wav", "wb") as f:
|
237 |
-
# f.write(audio_file.getbuffer())
|
238 |
-
|
239 |
-
# st.audio("test.wav", format="audio/wav")
|
240 |
-
|
241 |
-
# elif file_extension == ".mp3":
|
242 |
-
# with open("test.mp3", "wb") as f:
|
243 |
-
# f.write(audio_file.getbuffer())
|
244 |
-
|
245 |
-
# st.audio("test.mp3", format="audio/mpeg")
|
246 |
-
|
247 |
-
# #Load the WAV file using torchaudio
|
248 |
-
# if file_extension in [".wav", ".flac"]:
|
249 |
-
# wav, sample_rate = torchaudio.load("test.wav")
|
250 |
-
# # st.markdown("Before unsquueze wav")
|
251 |
-
# # st.markdown(wav)
|
252 |
-
# file_extension_ori =".wav"
|
253 |
-
# #Unsqueeze for line 176
|
254 |
-
# wav= wav.unsqueeze(0)
|
255 |
-
|
256 |
-
# elif file_extension == ".mp3":
|
257 |
-
# # Load an MP3 file
|
258 |
-
# audio = AudioSegment.from_mp3("test.mp3")
|
259 |
-
|
260 |
-
# # Export it as a WAV file
|
261 |
-
# audio.export("test.wav", format="wav")
|
262 |
-
# wav3, sample_rate = torchaudio.load("test.wav")
|
263 |
-
# wav= wav3.unsqueeze(0)
|
264 |
-
# file_extension_ori =".mp3"
|
265 |
-
# file_extension =".wav"
|
266 |
-
|
267 |
-
# #RuntimeError: Could not infer dtype of numpy.float32
|
268 |
-
# #wav = torch.tensor(wav3).float() / 32768.0
|
269 |
-
|
270 |
-
# #RuntimeError: Numpy is not available
|
271 |
-
# # wav = torch.from_numpy(wav3) #/32768.0
|
272 |
-
# # wav = wav.unsqueeze(0).unsqueeze(0)
|
273 |
-
# # st.markdown("Before unsqueeze mp3")
|
274 |
-
# # st.markdown(wav)
|
275 |
-
|
276 |
-
# #Unsqueeze for line 176
|
277 |
-
# # wav= wav.unsqueeze(0)
|
278 |
-
|
279 |
-
# action = st.selectbox("Select Action", ["Add Watermark", "Detect Watermark"])
|
280 |
-
|
281 |
-
# if action == "Add Watermark":
|
282 |
-
# #watermark_text = st.text_input("The watermark (0, 1 list of length-16):", value=st.session_state.def_value)
|
283 |
-
# add_watermark_button = st.button("Add Watermark", key="add_watermark_btn")
|
284 |
-
# if add_watermark_button: # 点击按钮后执行的
|
285 |
-
# #if audio_file and watermark_text:
|
286 |
-
# if audio_file:
|
287 |
-
# with st.spinner("Adding Watermark..."):
|
288 |
-
# #wav = my_read_file(wav,max_second_encode)
|
289 |
-
|
290 |
-
# #1st attempt
|
291 |
-
# watermark = model.get_watermark(wav, default_sr)
|
292 |
-
# watermarked_audio = wav + watermark
|
293 |
-
# print(watermarked_audio.size())
|
294 |
-
# size = watermarked_audio.size()
|
295 |
-
# #st.markdown(size)
|
296 |
-
|
297 |
-
# print(watermarked_audio.squeeze())
|
298 |
-
# squeeze = watermarked_audio.squeeze(1)
|
299 |
-
# shape = squeeze.size()
|
300 |
-
# #st.markdown(shape)
|
301 |
-
|
302 |
-
# #st.markdown(squeeze)
|
303 |
-
# if file_extension_ori in [".wav", ".flac"]:
|
304 |
-
# torchaudio.save("output.wav", squeeze, default_sr, bits_per_sample=16)
|
305 |
-
# watermarked_wav = torchaudio.save("output.wav", squeeze, default_sr, bits_per_sample=16)
|
306 |
-
|
307 |
-
# st.audio("output.wav", format="audio/wav")
|
308 |
-
|
309 |
-
# with open("output.wav", "rb") as file:
|
310 |
-
# #file.read()
|
311 |
-
# #file.write(watermarked_wav.getbuffer())
|
312 |
-
# binary_data = file.read()
|
313 |
-
# btn = st.download_button(
|
314 |
-
# label="Download watermarked audio",
|
315 |
-
# data=binary_data,
|
316 |
-
# file_name="output.wav",
|
317 |
-
# mime="audio/wav",
|
318 |
-
# )
|
319 |
-
|
320 |
-
|
321 |
-
# elif file_extension_ori == ".mp3":
|
322 |
-
# torchaudio.save("output.wav", squeeze, default_sr)
|
323 |
-
# watermarked_mp3 = torchaudio.save("output.wav", squeeze, default_sr)
|
324 |
-
# audio = AudioSegment.from_wav("output.wav")
|
325 |
-
|
326 |
-
# # Export as MP3
|
327 |
-
# audio.export("output.mp3", format="mp3")
|
328 |
-
# st.audio("output.mp3", format="audio/mpeg")
|
329 |
-
|
330 |
-
# with open("output.mp3", "rb") as file:
|
331 |
-
# #file.write(watermarked_wav.getbuffer())
|
332 |
-
# binary_data = file.read()
|
333 |
-
# st.download_button(
|
334 |
-
# label="Download watermarked audio",
|
335 |
-
# data=binary_data,
|
336 |
-
# file_name="output.mp3",
|
337 |
-
# mime="audio/mpeg",
|
338 |
-
# )
|
339 |
-
# # except RuntimeError:
|
340 |
-
# # st.error("Please input audio with one channel (mono-channel)")
|
341 |
-
|
342 |
-
# elif action == "Detect Watermark":
|
343 |
-
# detect_watermark_button = st.button("Detect Watermark", key="detect_watermark_btn")
|
344 |
-
|
345 |
-
# # if audio_file:
|
346 |
-
# # #1st attempt
|
347 |
-
# # watermark = model.get_watermark(wav, default_sr)
|
348 |
-
# # watermarked_audio = wav + watermark
|
349 |
-
# # print(watermarked_audio.size())
|
350 |
-
# # size = watermarked_audio.size()
|
351 |
-
# # #st.markdown(size)
|
352 |
-
|
353 |
-
|
354 |
-
# if detect_watermark_button:
|
355 |
-
# with st.spinner("Detecting..."):
|
356 |
-
# # result, message = detector.detect_watermark(watermarked_audio, sample_rate=default_sr, message_threshold=0.5)
|
357 |
-
# # st.markdown("Probability of audio being watermarked: ")
|
358 |
-
# # st.markdown(result)
|
359 |
-
# # st.markdown("This is likely a watermarked audio!")
|
360 |
-
# # print(f"\nThis is likely a watermarked audio: {result}")
|
361 |
-
|
362 |
-
# #Run on an unwatermarked audio
|
363 |
-
|
364 |
-
# if file_extension in [".wav", ".flac"]:
|
365 |
-
# wav, sample_rate = torchaudio.load("test.wav")
|
366 |
-
# wav= wav.unsqueeze(0)
|
367 |
-
|
368 |
-
# elif file_extension == ".mp3":
|
369 |
-
# # Load an MP3 file
|
370 |
-
# audio = AudioSegment.from_mp3("test.mp3")
|
371 |
-
# # Export it as a WAV file
|
372 |
-
# audio.export("test.wav", format="wav")
|
373 |
-
# wav, sample_rate = torchaudio.load("test.wav")
|
374 |
-
# wav= wav.unsqueeze(0)
|
375 |
-
|
376 |
-
# result2, message2 = detector.detect_watermark(wav, sample_rate=default_sr, message_threshold=0.5)
|
377 |
-
# print(f"This is likely an unwatermarked audio: {result2}")
|
378 |
-
# st.markdown("Probability of audio being watermarked: ")
|
379 |
-
# st.markdown(result2)
|
380 |
-
# st.markdown("This is likely an unwatermarked audio!")
|
381 |
|
382 |
|
383 |
if __name__ == "__main__":
|
@@ -389,9 +193,4 @@ if __name__ == "__main__":
|
|
389 |
# model = wavmark.load_model().to(device)
|
390 |
model = AudioSeal.load_generator("audioseal_wm_16bits")
|
391 |
detector = AudioSeal.load_detector(("audioseal_detector_16bits"))
|
392 |
-
main()
|
393 |
-
|
394 |
-
# audio_path = "/Users/my/Library/Mobile Documents/com~apple~CloudDocs/CODE/PycharmProjects/4_语音水印/419_huggingface水印/WavMark/example.wav"
|
395 |
-
|
396 |
-
# decoded_watermark, decode_cost = decode_watermark(audio_path)
|
397 |
-
# print(decoded_watermark)
|
|
|
14 |
import librosa
|
15 |
import ffmpeg
|
16 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
def create_default_value():
|
18 |
if "def_value" not in st.session_state:
|
19 |
def_val_npy = np.random.choice([0, 1], size=32 - len_start_bit)
|
|
|
87 |
file_extension_ori =".mp3"
|
88 |
file_extension =".wav"
|
89 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
action = st.selectbox("Select Action", ["Add Watermark", "Detect Watermark"])
|
91 |
|
92 |
if action == "Add Watermark":
|
|
|
182 |
|
183 |
except RuntimeError:
|
184 |
st.error("Please input audio with one channel (mono-channel)")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
185 |
|
186 |
|
187 |
if __name__ == "__main__":
|
|
|
193 |
# model = wavmark.load_model().to(device)
|
194 |
model = AudioSeal.load_generator("audioseal_wm_16bits")
|
195 |
detector = AudioSeal.load_detector(("audioseal_detector_16bits"))
|
196 |
+
main()
|
|
|
|
|
|
|
|
|
|