Update app.py

app.py

@@ -94,7 +94,7 @@ def predict_answer(video, image, question):
                 use_cache=True)[0]
         answer = tokenizer.decode(output_ids[input_ids.shape[1]:], skip_special_tokens=True).strip()
 
-        return ast.literal_eval(answer)
+        return answer
         # # Process as a video
         # frames = video_to_frames(video)
         # answers = []
@@ -158,8 +158,21 @@ promt_bus_people = """
 provide me the answers as a dictionary with key as the string value of the variable value on top and its value should be boolean value        
 """
 
+# promt_video = """
+#     Annotate this image with this schema:
+#     {
+#         “description”: “Is the person standing?”,
+#         “value”: “standing”,
+#     },
+#     {
+#         “description”: “Is the person's hands free?”,
+#         “value”: “Hands-Free”,
+#     },
+# provide me the answers as a dictionary with key as the string value of the variable value on top and its value should be boolean value
+# """
+
 promt_video = """
-    Annotate this image with this schema:
+    Annotate this image by indicating the presence or absence of specific objects.
     {
         “description”: “Is the person standing?”,
         “value”: “standing”,
@@ -168,8 +181,11 @@ promt_video = """
         “description”: “Is the person's hands free?”,
         “value”: “Hands-Free”,
     },
-provide me the answers as a dictionary with key as the string value of the variable value on top and its value should be boolean value
+Provide your answers as a dictionary with the object type as the key and a boolean value indicating its presence in the image, Use 'true' for objects present in the image and 'false' for objects not present.
+
 """
+
+
 test_examples = [[None, "Images/cat_dog.jpeg", promt_cat_dog], 
             [None,"Images/bus_people.jpeg", promt_bus_people], 
             ["videos/v2.mp4",None,promt_video],