Update app.py

app.py

@@ -50,9 +50,9 @@ def evaluate_caption(image, caption):
     score = similarity_score.softmax(dim=1).detach().numpy()
     print(score)
     if score[0][0]>score[0][1]:
-      winner = "The first caption is the human"
+      winner = "Player 1 wins!"
     else:
-      winner = "The second caption is the human"
+      winner = "Player 2 wins!"
 
     
     return blip_caption,winner
@@ -60,6 +60,14 @@ def evaluate_caption(image, caption):
 
 callback = gr.HuggingFaceDatasetSaver('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', "gradioTest")
 with gr.Blocks() as demo:
+  gr.Markdown(
+    """
+    # Welcome to our Human vs. AI game!
+    
+    ### You and an AI agent are trying to convince a third AI agent that each of you are better at describing the visual world. \n
+    ### In order to win, describe this image in one sentence. Then the second AI agent will also generate a description and the third agent will decide a winner.
+    ### You win if the AI says that "Player 1 wins!"
+    """)
   im_path_str = 'n01677366_12918.JPEG'
   im_path = gr.Textbox(label="Image fname",value=im_path_str,interactive=False, visible=False)
   # fn=evaluate_caption,