WIP: Options to use database annotations function instead of PadFile.annotations

Allows to use the database annotations function to get a file annotations.

bob/pad/base/pipelines/vanilla_pad/legacy.py

@@ -9,13 +9,15 @@ from .abstract_classes import Database
 logger = logging.getLogger(__name__)
 
 
-def _padfile_to_delayed_sample(padfile, database):
+def _padfile_to_delayed_sample(padfile, database, use_db_annotations):
     return DelayedSample(
         load=padfile.load,
         subject=str(padfile.client_id),
         attack_type=padfile.attack_type,
         key=padfile.path,
-        delayed_attributes=dict(annotations=lambda : padfile.annotations),
+        delayed_attributes=dict(
+            annotations=lambda: database.annotations(padfile) if use_db_annotations else padfile.annotations
+        ),
         is_bonafide=padfile.attack_type is None,
     )
 
@@ -34,16 +36,17 @@ class DatabaseConnector(Database):
     """
 
     def __init__(
-        self, database, annotation_type="eyes-center", fixed_positions=None, **kwargs
+        self, database, annotation_type="eyes-center", use_db_annotations=False, fixed_positions=None, **kwargs
     ):
         super().__init__(**kwargs)
         self.database = database
         self.annotation_type = annotation_type
+        self.use_db_annotations = use_db_annotations
         self.fixed_positions = fixed_positions
 
     def fit_samples(self):
         objects = self.database.training_files(flat=True)
-        return [_padfile_to_delayed_sample(k, self.database) for k in objects]
+        return [_padfile_to_delayed_sample(k, self.database, self.use_db_annotations) for k in objects]
 
     def predict_samples(self, group="dev"):
         objects = self.database.all_files(groups=group, flat=True)