results

.gitignore

@@ -1 +1,3 @@
-__pycache__
+__pycache__
+venv
+gradio_queue*

app.py

@@ -46,11 +46,13 @@ def inference(audio):
     result = model.inference(in_val)
     pred = result['clipwise_output'][0]
     # pred = np.exp(pred)/np.sum(np.exp(pred)) # softmax
-    return {class_mapping[i]: float(p) for i, p in enumerate(pred)}
-    # win_classes = np.argmax(result['clipwise_output'], axis=1)
-    # win_class_index = win_classes[0]
-    # win_class_name = class_mapping[win_class_index]
+    output = {class_mapping[i]: float(p) for i, p in enumerate(pred)}
+    
+    win_classes = np.argmax(result['clipwise_output'], axis=1)
+    win_class_index = win_classes[0]
+    win_class_name = class_mapping[win_class_index]
     # return str({win_class_name: result['clipwise_output'][0][win_class_index]})
+    return str({win_class_name: result['clipwise_output'][0][win_class_index]}), output
 
 
 title = "HTS-Audio-Transformer"
@@ -62,7 +64,7 @@ gr.Interface(
     inference,
     gr.inputs.Audio(type="numpy", label="Input"),
     # gr.outputs.Textbox(),
-    gr.outputs.Label(),
+    [gr.outputs.Textbox(), gr.outputs.JSON()],
     title=title,
     description=description,
     # article=article,