Update app.py

app.py

@@ -40,8 +40,9 @@ DESCRIPTION = '''# ECCV'22 | Panoptic Scene Graph Generation
 
 🎯 🎯 🎯 The PSG Development Team is currently focusing on **(1) 🧙‍♂️ Next-Generation PSG Models**, **(2) 🕵️‍♀️ Relation-Aware Visual Reasoning from PSG Models**, and **(3) 🎨 Relation-Aware Image Generation from Scene Graph and Caption**. If you are also interested in the related researches, please reach out and contact us!
 
-Inference takes 10-30 seconds per image. The model is PSGTR (60 epochs). You can upload your own pictures to play, or select the examples below.
-The demo will output a GIF to show the first 10 "subject-verb-object" relations, and a gallery attached below.
+Inference takes 10-30 seconds per image. The model is PSGTR (60 epochs). You can upload your own pictures or select the examples below to play.
+The demo will output a GIF to show the first 10 "subject-verb-object" relations, with the subject and object being grounded by segmentation masks. 
+A gallery is attached below for reference.
 
 '''
 FOOTER = '<img id="visitor-badge" src="https://visitor-badge.glitch.me/badge?page_id=c-liangyu.openpsg" alt="visitor badge" />'