c

app.py

@@ -12,6 +12,7 @@ model.to(device)
 
 def classify_sound(file_path):
     wv, sr = torchaudio.load(file_path)
+    original_shape = wv.shape
 
     # Convert to mono
     if wv.shape[0] > 1:
@@ -27,16 +28,33 @@ def classify_sound(file_path):
     probs = torch.softmax(logits, dim=-1)[0]
     top5 = torch.topk(probs, k=5)
 
-    res = [
+    top5_labels = [
         (model.config.id2label[idx.item()], round(prob.item(), 4))
         for idx, prob in zip(top5.indices, top5.values)
     ]
-    return dict(res)
+
+    full_probs = {
+        model.config.id2label[i]: round(prob.item(), 4)
+        for i, prob in enumerate(probs)
+    }
+
+    return {
+        "Top 5 Predictions": dict(top5_labels),
+        "Sampling Rate": sr,
+        "Waveform Shape": str(original_shape),
+        "All Probabilities": full_probs
+    }
+
 
 demo = gr.Interface(
     fn=classify_sound, 
     inputs=gr.Audio(sources="upload", type="filepath"),
-    outputs=gr.Label(num_top_classes=5),
+    outputs=[
+        gr.Label(label = "Top 5 Pred", num_top_classes=5),
+        gr.Textbox(label="Sample Rate"),
+        gr.Textbox(label="Waveform Shape"),
+        gr.JSON(label="All Class Probabilities")
+    ],
     title="Audio Classification with AST",
     description="Upload an audio clip (speech, music, ambient sound, etc.). Model: MIT AST fine-tuned on AudioSet (10 classes).",
     live=False,