Update app.py

add description by llm

app.py

@@ -1,4 +1,4 @@
-from utils.utils import get_logger, initialization, get_result
+from utils.utils import get_logger, initialization, get_search_result, get_image_description
 import gradio as gr
 import logging
 
@@ -15,9 +15,6 @@ def main(query):
     logger.info("-------------------------------------------------------")
     exit = False
     while not exit:
-        # Collect user query
-        # query = input('Type your query, or "exit" if you want to exit: ')
-
         if query == "exit":
             exit = True
             print("-------------------------------------------------------")
@@ -26,24 +23,26 @@ def main(query):
             logger.info("Search terminated.")
             return None, "Search terminated."
         else:
-            # Get search result including the original descriptions of the images
-            image, text = get_result(collection, data_set, query, model, n_results=2)
+            # Get search result including the original description of the image
+            image, text = get_search_result(collection, data_set, query, model, n_results=2)
 
-            # Display the image, its caption, and user query
-            # show_image(image, text, query)
-            return image, text
+            # Get detailed description of the image generated by multimodal LLM
+            description_by_llm = get_image_description(image)
+            
+            return image, text, description_by_llm
 
 
 if __name__ == "__main__":
     try:
         if collection == None:
             collection, data_set, model, logger = initialization(logger)
-        # main()
+        
         app = gr.Interface(
             fn=main,
             inputs=[gr.Textbox(label="Describe the scene that you are looking for:")],
-            outputs=[gr.Image(label="Here's the scene found based on your description:"),
-                     gr.Textbox(label="Original description of the found scene:")],
+            outputs=[gr.Image(label="Here's the scene found based on your input:"),
+                     gr.Textbox(label="Original description of the found scene:"),
+                     gr.Textbox(label="Detailed description generated by LLM:")],
             title="Search for a scene in the world of GTA!"
         )
         app.launch(share=True)