Fix bug in image loader v3

app.py

@@ -62,7 +62,7 @@ def greet(image, audio):
 
     # Localization result
     out_dict = model(image_file.to(model.device), audio_driven_embedding, 352)
-    seg = out_dict['heatmap'][j:j + 1]
+    seg = out_dict['heatmap'][0:1]
     seg_image = ((1 - seg.squeeze().detach().cpu().numpy()) * 255).astype(np.uint8)
     seg_image = Image.fromarray(seg_image)
     heatmap_image = cv2.applyColorMap(np.array(seg_image), cv2.COLORMAP_JET)
@@ -70,8 +70,12 @@ def greet(image, audio):
 
     return overlaid_image
 
-title = 'Audio-Grounded Contrastive Learning (ACL)'
-description = "This is your private demo of our WACV'24 paper Can CLIP Help Sound Source Localization?"
+
+title = "Zero-shot sound source localization with ACL"
+description = "This is simple demo of our WACV'24 paper 'Can CLIP Help Sound Source Localization?'\n"
+description += "To use it, simply upload an image and corresponding audio to mask (identify in the image)"
+description += "or use one of the examples below and click 'submit'. Results will show up in a few seconds."
+
 
 demo = gr.Interface(
     fn=greet,