Create app.py

app.py

@@ -0,0 +1,50 @@
+import gradio as gr
+import torchaudio
+from transformers import pipeline
+from datasets import load_dataset, Audio
+
+# Load your model
+classifier = pipeline("audio-classification", model="candenizkocak/wav2vec2-base_turkish_gender_classification")
+
+# Function to resample audio to 16kHz
+def resample_audio(audio_file, target_sampling_rate=16000):
+    waveform, original_sample_rate = torchaudio.load(audio_file)
+    if original_sample_rate != target_sampling_rate:
+        resampler = torchaudio.transforms.Resample(orig_freq=original_sample_rate, new_freq=target_sampling_rate)
+        waveform = resampler(waveform)
+    return waveform.squeeze().numpy(), target_sampling_rate
+
+# Define the prediction function
+def classify_audio(audio_file):
+    # Resample the audio to 16kHz
+    resampled_audio, _ = resample_audio(audio_file)
+    
+    # Classify the audio
+    prediction = classifier(resampled_audio)
+    
+    # Return predictions as a dictionary
+    return {entry['label']: entry['score'] for entry in prediction}
+
+# Define Gradio interface
+def demo():
+    with gr.Blocks(theme=gr.themes.Soft()) as demo:
+        gr.Markdown("## Turkish Gender Audio Classification")
+        
+        # Input Audio
+        with gr.Row():
+            audio_input = gr.Audio(type="filepath", label="Input Audio")
+        
+        # Output Labels
+        with gr.Row():
+            label_output = gr.Label(label="Prediction")
+        
+        # Predict Button
+        classify_btn = gr.Button("Classify")
+
+        # Define the interaction
+        classify_btn.click(fn=classify_audio, inputs=audio_input, outputs=label_output)
+        
+    return demo
+
+# Launch the demo
+demo().launch()