Update app.py

app.py

@@ -5,38 +5,44 @@ from jiwer import wer
 
 # Load models
 whisper_pipeline_1 = pipeline("automatic-speech-recognition", model="maliahson/Finetuned_Whisper_Medium_Model_2")
-device = 0 if torch.cuda.is_available() else "cpu"
-whisper_pipeline_2 = pipeline("automatic-speech-recognition", model="openai/whisper-large-v3-turbo", device=device)
+whisper_pipeline_2 = pipeline("automatic-speech-recognition", model="openai/whisper-large-v3-turbo", device=0 if torch.cuda.is_available() else "cpu")
+whisper_pipeline_3 = pipeline("automatic-speech-recognition", model="ihanif/whisper-medium-urdu", device=0 if torch.cuda.is_available() else "cpu")
 
 def transcribe_and_compare(audio_path, original_transcription=None):
     """
-    Transcribes an audio file using two Whisper models and compares results.
+    Transcribes an audio file using three Whisper models and compares results.
     
     Args:
         audio_path (str): Path to the audio file.
         original_transcription (str, optional): Ground truth transcription.
-
     Returns:
         dict: Results including transcriptions and WER calculations.
     """
     transcription_1 = whisper_pipeline_1(audio_path)["text"]
     transcription_2 = whisper_pipeline_2(audio_path)["text"]
+    transcription_3 = whisper_pipeline_3(audio_path)["text"]
+    
     comparison_result = {
-        "Model 1 Output (maliahson/whisper-agri)": transcription_1,
+        "Model 1 Output (maliahson/Finetuned_Whisper_Medium_Model_2)": transcription_1,
         "Model 2 Output (openai/whisper-large-v3-turbo)": transcription_2,
+        "Model 3 Output (ihanif/whisper-medium-urdu)": transcription_3
     }
 
     if original_transcription:
-        # Calculate Word Error Rate
+        # Calculate Word Error Rate (WER)
         wer_1 = wer(original_transcription, transcription_1)
         wer_2 = wer(original_transcription, transcription_2)
+        wer_3 = wer(original_transcription, transcription_3)
+        
         comparison_result["WER Model 1"] = wer_1
         comparison_result["WER Model 2"] = wer_2
+        comparison_result["WER Model 3"] = wer_3
     else:
-        # Compare outputs of both models
+        # Compare outputs of all three models
         comparison_result["Difference Between Models"] = {
-            "Model 1 Unique Words": set(transcription_1.split()) - set(transcription_2.split()),
-            "Model 2 Unique Words": set(transcription_2.split()) - set(transcription_1.split()),
+            "Model 1 Unique Words": set(transcription_1.split()) - set(transcription_2.split()) - set(transcription_3.split()),
+            "Model 2 Unique Words": set(transcription_2.split()) - set(transcription_1.split()) - set(transcription_3.split()),
+            "Model 3 Unique Words": set(transcription_3.split()) - set(transcription_1.split()) - set(transcription_2.split()),
         }
 
     return comparison_result
@@ -55,4 +61,4 @@ with gr.Blocks() as demo:
         outputs=output
     )
 
-demo.launch(debug=True)
+demo.launch(debug=True)