Update dataset

ovdino/detrex/data/datasets/coco_ovd.py

@@ -25,6 +25,7 @@
 
 
 logger = logging.getLogger(__name__)
+print_sample = True
 
 __all__ = ["load_coco_json", "register_coco_ovd_instances"]
 
@@ -73,6 +74,7 @@ def load_coco_json(
            The results do not have the "image" field.
     """
     timer = Timer()
+    global print_sample
     json_file = PathManager.get_local_path(json_file)
     with contextlib.redirect_stdout(io.StringIO()):
         coco_api = COCO(json_file)
@@ -238,23 +240,20 @@ def load_coco_json(
         record["annotations"] = objs
 
         if test_mode and num_sampled_classes > 0:
+            assert template == "identity"
             obj_cat_ids = [obj["category_id"] for obj in objs]
             sampled_cat_names = [
                 clean_words_or_phrase(cat_name) for _, cat_name in id2name.items()
             ]
-            sampled_cat_names = [
-                [template.format(cat_name) for template in template_meta[template]]
-                for cat_name in sampled_cat_names
-            ]
 
         # sample category from category_list
         if not test_mode and num_sampled_classes > 0:
             obj_cat_ids = [obj["category_id"] for obj in objs]
-            continous_cat_ids = sorted(id_map.values())
+            continuous_cat_ids = sorted(id_map.values())
             pos_cat_ids = set(obj_cat_ids)
             assert len(pos_cat_ids) <= num_sampled_classes
             neg_cat_ids = random.sample(
-                set(continous_cat_ids) - pos_cat_ids,
+                set(continuous_cat_ids) - pos_cat_ids,
                 num_sampled_classes - len(pos_cat_ids),
             )
             sampled_cat_ids = list(pos_cat_ids) + list(neg_cat_ids)
@@ -275,10 +274,12 @@ def load_coco_json(
 
         dataset_dicts.append(record)
 
-    rank0_print(
-        f"Loaded {len(dataset_dicts)} data points from {dataset_name}, template: {template}\n"
-        + f"Sample: {sampled_cat_names}"
-    )
+    if print_sample:
+        rank0_print(
+            f"Loaded {len(dataset_dicts)} data points from {dataset_name}, template: {template}\n"
+            + f"Sample: {sampled_cat_names}"
+        )
+        print_sample = False
 
     if num_instances_without_valid_segmentation > 0:
         logger.warning(

ovdino/detrex/data/datasets/custom_ovd.py

@@ -25,6 +25,7 @@
 
 
 logger = logging.getLogger(__name__)
+print_sample = True
 
 __all__ = ["load_custom_json", "register_custom_ovd_instances"]
 
@@ -73,6 +74,7 @@ def load_custom_json(
            The results do not have the "image" field.
     """
     timer = Timer()
+    global print_sample
     json_file = PathManager.get_local_path(json_file)
     with contextlib.redirect_stdout(io.StringIO()):
         coco_api = COCO(json_file)
@@ -238,23 +240,20 @@ def load_custom_json(
         record["annotations"] = objs
 
         if test_mode and num_sampled_classes > 0:
+            assert template == "identity"
             obj_cat_ids = [obj["category_id"] for obj in objs]
             sampled_cat_names = [
                 clean_words_or_phrase(cat_name) for _, cat_name in id2name.items()
             ]
-            sampled_cat_names = [
-                [template.format(cat_name) for template in template_meta[template]]
-                for cat_name in sampled_cat_names
-            ]
 
         # sample category from category_list
         if not test_mode and num_sampled_classes > 0:
             obj_cat_ids = [obj["category_id"] for obj in objs]
-            continous_cat_ids = sorted(id_map.values())
+            continuous_cat_ids = sorted(id_map.values())
             pos_cat_ids = set(obj_cat_ids)
             assert len(pos_cat_ids) <= num_sampled_classes
             neg_cat_ids = random.sample(
-                set(continous_cat_ids) - pos_cat_ids,
+                set(continuous_cat_ids) - pos_cat_ids,
                 num_sampled_classes - len(pos_cat_ids),
             )
             sampled_cat_ids = list(pos_cat_ids) + list(neg_cat_ids)
@@ -275,10 +274,12 @@ def load_custom_json(
 
         dataset_dicts.append(record)
 
-    rank0_print(
-        f"Loaded {len(dataset_dicts)} data points from {dataset_name}, template: {template}\n"
-        + f"Sample: {sampled_cat_names}"
-    )
+    if print_sample:
+        rank0_print(
+            f"Loaded {len(dataset_dicts)} data points from {dataset_name}, template: {template}\n"
+            + f"Sample: {sampled_cat_names}"
+        )
+        print_sample = False
 
     if num_instances_without_valid_segmentation > 0:
         logger.warning(