Fix bug in image loader

app.py

@@ -53,7 +53,7 @@ def greet(image, audio):
         vt.Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)),  # CLIP
     ])
 
-    image_file = image_transform(image)
+    image_file = image_transform(image).unsqeeze(0)
 
     # Inference
     placeholder_tokens = model.get_placeholder_token(prompt_template.replace('{}', ''))
@@ -70,8 +70,8 @@ def greet(image, audio):
 
     return overlaid_image
 
-
-description = 'hello world'
+title = 'Audio-Grounded Contrastive Learning (ACL)'
+description = "This is your private demo of our WACV'24 paper Can CLIP Help Sound Source Localization?"
 
 demo = gr.Interface(
     fn=greet,