changes3

app.py

@@ -9,6 +9,7 @@ import matplotlib.pyplot as plt
 import matplotlib
 matplotlib.use('Agg')  # Use non-interactive backend
 from io import BytesIO
+from PIL import Image
 import warnings
 
 # Suppress warnings and logs
@@ -69,12 +70,13 @@ def create_mel_spectrogram(audio, sr):
     plt.title('Mel Spectrogram')
     plt.tight_layout()
     
-    # Save to BytesIO and return the plot
+    # Save to BytesIO and convert to PIL Image
     buf = BytesIO()
     plt.savefig(buf, format='png', dpi=150, bbox_inches='tight')
     buf.seek(0)
+    img = Image.open(buf)
     plt.close()
-    return buf
+    return img
 
 def create_polar_plot(emotion_probabilities):
     """Create polar plot of emotion probabilities"""
@@ -101,12 +103,13 @@ def create_polar_plot(emotion_probabilities):
     ax.set_title("Emotion Probabilities", va='bottom', fontsize=14, color="darkblue", pad=20)
     plt.tight_layout()
     
-    # Save to BytesIO and return the plot
+    # Save to BytesIO and convert to PIL Image
     buf = BytesIO()
     plt.savefig(buf, format='png', dpi=150, bbox_inches='tight')
     buf.seek(0)
+    img = Image.open(buf)
     plt.close()
-    return buf
+    return img
 
 def create_waveform_plot(audio, sr):
     """Create waveform plot"""
@@ -117,12 +120,13 @@ def create_waveform_plot(audio, sr):
     plt.ylabel('Amplitude')
     plt.tight_layout()
     
-    # Save to BytesIO and return the plot
+    # Save to BytesIO and convert to PIL Image
     buf = BytesIO()
     plt.savefig(buf, format='png', dpi=150, bbox_inches='tight')
     buf.seek(0)
+    img = Image.open(buf)
     plt.close()
-    return buf
+    return img
 
 def predict_emotion(audio_file):
     try:
@@ -213,6 +217,21 @@ with gr.Blocks(title="🎤 Emotion Recognition from Audio", theme=gr.themes.Soft
         outputs=[predicted_emotion, emotion_probs, mel_spec_plot, polar_plot, waveform_plot]
     )
 
+    gr.Markdown(
+        """
+        ### 📝 How it works:
+        1. **Upload** an audio file or **record** directly using your microphone
+        2. The system extracts audio features (MFCCs, Chroma, Spectral features, etc.)
+        3. A trained neural network predicts the emotion
+        4. View the results with detailed visualizations:
+           - **Waveform**: Shows the audio signal over time
+           - **Mel Spectrogram**: Visual representation of the audio's frequency content
+           - **Radar Chart**: Probability distribution across all emotion categories
+        
+        ### 🎭 Supported Emotions:
+        Depending on your model training, this may include emotions like: Happy, Sad, Angry, Fear, Disgust, Surprise, Neutral, and others.
+        """
+    )
 
 # Launch the interface
 if __name__ == "__main__":