commit

app.py

@@ -18,7 +18,7 @@ bert_atco_ner = pipeline(model='Jzuluaga/bert-base-ner-atc-en-atco2-1h')
 whisper_v2 = pipeline(model='jlvdoorn/whisper-large-v2-atco2-asr-atcosim')
 whisper_v3 = pipeline(model='jlvdoorn/whisper-large-v3-atco2-asr-atcosim')
 #%%
-def transcribe(audio_file, audio_mic, model_version):
+def transcribe(audio, model_version):
     if model_version == 'large-v2':
         whisper = whisper_v2
         ttl = 'Whisper Large v2 - ATCO2-ATCOSIM'
@@ -27,10 +27,8 @@ def transcribe(audio_file, audio_mic, model_version):
         whisper = whisper_v3
         ttl = 'Whisper Large v3 - ATCO2-ATCOSIM'
         dis = 'This demo will transcribe ATC audio files by using the Whisper Large v3 model fine-tuned on the ATCO2 and ATCOSIM datasets. \n \n Further it uses a Named Entity Recognition model to extract callsigns, commands and values from the transcription. \n This model is based on Google\'s BERT model and fine-tuned on the ATCO2 dataset.'
-    if audio_mic is not None:
-        return whisper(audio_mic)['text']
-    elif audio_file is not None:
-        return whisper(audio_file)['text']
+    if audio is not None:
+        return whisper(audio)['text']
     else:
         return 'There was no audio to transcribe...'
 
@@ -54,8 +52,8 @@ def extractCallSignCommand(transcription):
         return 'There was no transcription to extract a callsign or command from...'
 
 #%%
-def transcribeAndExtract(audio_file, audio_mic, transcribe_only, model_version):
-    transcription = transcribe(audio_file, audio_mic, model_version)
+def transcribeAndExtract(audio, transcribe_only, model_version):
+    transcription = transcribe(audio, model_version)
     if not transcribe_only:
         callSignCommandValues = extractCallSignCommand(transcription)
     else:
@@ -79,5 +77,29 @@ iface = gr.Interface(
 )
 
 #%%
-#iface.launch(server_name='0.0.0.0', server_port=9000)
-iface.launch()
+file_iface = gr.load(
+    fn = transcribeAndExtract,
+    inputs = [gr.Audio(source='upload', type='filepath', interactive=True), 
+              gr.Checkbox(label='Transcribe only', default=False), 
+              gr.Dropdown(choices=['large-v2', 'large-v3'], value='large-v3', label='Whisper model version')
+            ],
+    
+    outputs = [gr.Textbox(label='Transcription'), gr.Textbox(label='Callsigns, commands and values')],
+    title = 'Whisper ATC - Large v3',
+    description = 'Transcribe and extract',
+)
+
+mic_iface = gr.load(
+    fn = transcribeAndExtract,
+    inputs = [gr.Audio(source='microphone', type='filepath'), 
+              gr.Checkbox(label='Transcribe only', default=False), 
+              gr.Dropdown(choices=['large-v2', 'large-v3'], value='large-v3', label='Whisper model version')
+            ],
+    
+    outputs = [gr.Textbox(label='Transcription'), gr.Textbox(label='Callsigns, commands and values')],
+    title = 'Whisper ATC - Large v3',
+    description = 'Transcribe and extract',
+)
+#%%
+demo = gr.TabbedInterface([file_iface, mic_iface], ["File", "Microphone"])
+demo.launch()