Update app.py

app.py

@@ -5,63 +5,19 @@ from torch.nn.functional import softmax
 import librosa
 import os
 
-# Path to the local directory where the model files are stored
 local_model_path = "./"
-
-# Load the model and feature extractor outside the function to improve performance
 extractor = AutoFeatureExtractor.from_pretrained(local_model_path)
 model = AutoModelForAudioClassification.from_pretrained(local_model_path)
 
-def safe_path_join(base_path, path):
-    """
-    Safely join a base path and a potentially unsafe relative path.
-    
-    Args:
-        base_path: The base directory path.
-        path: The relative path to join with the base path.
-    
-    Returns:
-        The safely joined path if it's a subpath of the base_path, otherwise None.
-    """
-    base_path = os.path.abspath(os.path.normpath(base_path))
-    target_path = os.path.abspath(os.path.normpath(os.path.join(base_path, path)))
-    if os.path.commonpath([base_path]) == os.path.commonpath([base_path, target_path]):
-        return target_path
-    else:
-        return None
-
 def preprocess_audio(audio_file_path, target_sample_rate=16000):
-    """
-    Preprocesses the audio file for compatibility with the model's expectations.
-    
-    Args:
-        audio_file_path: Path to the audio file.
-        target_sample_rate: Desired sample rate compatible with the model.
-    
-    Returns:
-        Processed waveform and sample rate.
-    """
     waveform, _ = librosa.load(audio_file_path, sr=target_sample_rate, mono=True)
     return waveform, target_sample_rate
 
 def predict_voice(audio_file_path):
-    """
-    Predicts whether a voice is real or spoofed from an audio file.
-    
-    Args:
-        audio_file_path: The path to the input audio file to be classified.
-    
-    Returns:
-        A string with the prediction and confidence level.
-    """
-    expected_base_path = "/expected/path/for/safety"
-    safe_audio_file_path = safe_path_join(expected_base_path, audio_file_path)
-    
-    if not safe_audio_file_path:
-        return "Error: Invalid file path."
-    
     try:
-        waveform, sample_rate = preprocess_audio(safe_audio_file_path)
+        # In Hugging Face Spaces, uploaded files are temporarily stored in a way that's accessible
+        # to the app, so there's no need for a strict path check here.
+        waveform, sample_rate = preprocess_audio(audio_file_path)
         inputs = extractor(waveform, return_tensors="pt", sampling_rate=sample_rate)
         
         with torch.no_grad():
@@ -78,15 +34,12 @@ def predict_voice(audio_file_path):
     
     return result
 
-# Initialize Gradio interface without the enable_queue parameter
 iface = gr.Interface(
     fn=predict_voice,
     inputs=gr.Audio(label="Upload Audio File", type="filepath"),
     outputs=gr.Textbox(label="Prediction"),
     title="Voice Authenticity Detection",
-    description="Detects whether a voice is real or AI-generated. Upload an audio file to see the results.",
-    theme="huggingface"
+    description="Detects whether a voice is real or AI-generated. Upload an audio file to see the results."
 )
 
-# Launch the Gradio app
-iface.launch(share=True)
+iface.launch()