backup

app.py

@@ -182,10 +182,31 @@ def get_annotated_indices(username):
         return set()
 
 
+def get_annotated_post_ids(username):
+    """Get list of post_ids already annotated by this user"""
+    filename = "data/evaluation_results_exp.csv"
+    if not os.path.exists(filename):
+        print(f"No annotations found for user {username} (file doesn't exist)")
+        return set()
+
+    try:
+        df = pd.read_csv(filename)
+        if "post_id" not in df.columns or "username" not in df.columns:
+            print(f"No annotations found for user {username} (missing columns)")
+            return set()
+        user_annotations = df[df["username"] == username]["post_id"].tolist()
+        print(f"User {username} has seen {len(set(user_annotations))} unique posts")
+        return set(user_annotations)
+    except:
+        print(f"Error reading annotations for user {username}")
+        return set()
+
+
 def get_random_sample(username):
-    """Get a random sample excluding previously annotated items"""
-    # Get indices already annotated by this user
+    """Get a random sample trying to avoid previously seen post_ids"""
+    # Get indices and post_ids already annotated by this user
     annotated_indices = get_annotated_indices(username)
+    annotated_post_ids = get_annotated_post_ids(username)
 
     # Get all valid indices that haven't been annotated
     all_indices = set(range(len(dataset)))
@@ -195,9 +216,18 @@ def get_random_sample(username):
         # If user has annotated all items, allow repeats
         available_indices = list(all_indices)
 
-    # Randomly select from available indices
-    idx = random.choice(available_indices)
-    sample = dataset[idx]
+    # Try up to 5 times to get a sample with unseen post_id
+    max_attempts = 5
+    for _ in range(max_attempts):
+        idx = random.choice(available_indices)
+        sample = dataset[idx]
+        if sample["post_id"] not in annotated_post_ids:
+            break
+        # Remove this index from available indices for next attempt
+        available_indices.remove(idx)
+        if not available_indices:
+            # If no more indices available, use the last sampled one
+            break
 
     # Randomly decide which image goes to position A and B
     if random.choice([True, False]):
@@ -226,7 +256,7 @@ def get_random_sample(username):
         "image_b": image_b,
         "model_a": model_a,
         "model_b": model_b,
-        "dataset_idx": idx,  # Include the dataset index in the return
+        "dataset_idx": idx,
     }