Update app.py

app.py

@@ -54,13 +54,21 @@ def process_video_audio(video_path, audio_path):
             frame_idx += 1
     cap.release()
 
+    train_visual = tf.convert_to_tensor(train_visual, dtype=tf.float16)
+    train_audio_wave = tf.reshape(tf.convert_to_tensor(train_audio_wave, dtype=tf.float16), (1, 20, 13077))
+    train_audio_cnn = tf.convert_to_tensor(train_audio_cnn, dtype=tf.float16)
+
     return train_visual, train_audio_wave, train_audio_cnn
 
 # Định nghĩa giao diện Gradio
 def gradio_interface(video, audio):
+    
     train_visual, train_audio_wave, train_audio_cnn = process_video_audio(video, audio)
     model = trained_model("./model_vui_ve.h5")
-    return frame
+    output = model.predict({"input_visual": train_visual, "input_audio_cnn": train_audio_cnn, "input_audio_wave": train_audio_wave})
+    emo_index = tf.math.argmax(output)
+    
+    return emotion_labels[emo_index]
 
 iface = gr.Interface(
     fn=gradio_interface,
@@ -69,7 +77,7 @@ iface = gr.Interface(
         gr.Audio()
     ],
     outputs=[
-        gr.Image()
+        gr.Text()
     ],
     live=True,
     title="Video and Audio Processing with Emotion Recognition"