cdactvm commited on
Commit
ad2d740
·
verified ·
1 Parent(s): 3137a68

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +10 -10
app.py CHANGED
@@ -29,7 +29,7 @@ from wienerFilter import wiener_filter
29
  from highPassFilter import high_pass_filter
30
  from waveletDenoise import wavelet_denoise
31
 
32
- transcriber_hindi_new = pipeline(task="automatic-speech-recognition", model="cdactvm/w2v-bert-tamil_new")
33
  # transcriber_hindi_old = pipeline(task="automatic-speech-recognition", model="cdactvm/w2v-bert-tamil_new")
34
  processor = AutoProcessor.from_pretrained("cdactvm/w2v-bert-tamil_new")
35
  vocab_dict = processor.tokenizer.get_vocab()
@@ -49,7 +49,7 @@ processor.feature_extractor._processor_class = "Wav2Vec2ProcessorWithLM"
49
 
50
  def transcribe_tamil_new(audio):
51
  # # Process the audio file
52
- transcript = transcriber_hindi_new(audio)
53
  text_value = transcript['text']
54
  processd_doubles=process_doubles(text_value)
55
  replaced_words = replace_words(processd_doubles)
@@ -83,7 +83,7 @@ def Noise_cancellation_function(audio_file):
83
  write(temp_wav, sr, denoised_audio)
84
 
85
  # Perform speech recognition on the denoised audio
86
- transcript = transcriber_hindi_lm(temp_wav)
87
  text_value = transcript['text']
88
  cleaned_text=text_value.replace("<s>","")
89
  processd_doubles=process_doubles(cleaned_text)
@@ -102,13 +102,13 @@ def sel_lng(lng, mic=None, file=None):
102
  return "You must either provide a mic recording or a file"
103
 
104
  if lng == "model_1":
105
- return transcribe_hindi_old(audio)
106
- elif lng == "model_2":
107
- return transcribe_hindi_new(audio)
108
- elif lng== "model_3":
109
- return transcribe_hindi_lm(audio)
110
- elif lng== "model_4":
111
- return Noise_cancellation_function(audio)
112
 
113
 
114
  demo=gr.Interface(
 
29
  from highPassFilter import high_pass_filter
30
  from waveletDenoise import wavelet_denoise
31
 
32
+ transcriber_taml_new = pipeline(task="automatic-speech-recognition", model="cdactvm/w2v-bert-tamil_new")
33
  # transcriber_hindi_old = pipeline(task="automatic-speech-recognition", model="cdactvm/w2v-bert-tamil_new")
34
  processor = AutoProcessor.from_pretrained("cdactvm/w2v-bert-tamil_new")
35
  vocab_dict = processor.tokenizer.get_vocab()
 
49
 
50
  def transcribe_tamil_new(audio):
51
  # # Process the audio file
52
+ transcript = transcriber_taml_new(audio)
53
  text_value = transcript['text']
54
  processd_doubles=process_doubles(text_value)
55
  replaced_words = replace_words(processd_doubles)
 
83
  write(temp_wav, sr, denoised_audio)
84
 
85
  # Perform speech recognition on the denoised audio
86
+ transcript = transcriber_taml_new(temp_wav)
87
  text_value = transcript['text']
88
  cleaned_text=text_value.replace("<s>","")
89
  processd_doubles=process_doubles(cleaned_text)
 
102
  return "You must either provide a mic recording or a file"
103
 
104
  if lng == "model_1":
105
+ return transcriber_taml_new(audio)
106
+ # elif lng == "model_2":
107
+ # return transcribe_hindi_new(audio)
108
+ # elif lng== "model_3":
109
+ # return transcribe_hindi_lm(audio)
110
+ # elif lng== "model_4":
111
+ # return Noise_cancellation_function(audio)
112
 
113
 
114
  demo=gr.Interface(