set FILTER_EMPTY_ANNOTATIONS to False for panoptic segmentation

Summary:
Pull Request resolved: fairinternal/detectron2#429

In the past we don't have this config option, so we use some heuristics in code.
It's better to set it explicitly now that we have an option.

Reviewed By: alexander-kirillov, bowenc0221

Differential Revision: D22370318

fbshipit-source-id: 269fdfbf146263ba416424a24da0ad987acd4475

configs/COCO-PanopticSegmentation/Base-Panoptic-FPN.yaml

@@ -7,3 +7,5 @@ MODEL:
 DATASETS:
   TRAIN: ("coco_2017_train_panoptic_separated",)
   TEST: ("coco_2017_val_panoptic_separated",)
+DATALOADER:
+  FILTER_EMPTY_ANNOTATIONS: False

detectron2/config/defaults.py

@@ -117,8 +117,8 @@
 _C.DATALOADER.SAMPLER_TRAIN = "TrainingSampler"
 # Repeat threshold for RepeatFactorTrainingSampler
 _C.DATALOADER.REPEAT_THRESHOLD = 0.0
-# if True, the dataloader will filter out images that have no associated
-# annotations at train time.
+# Tf True, when working on datasets that have instance annotations, the
+# training dataloader will filter out images without associated annotations
 _C.DATALOADER.FILTER_EMPTY_ANNOTATIONS = True
 
 # ---------------------------------------------------------------------------- #

detectron2/data/build.py

@@ -211,6 +211,9 @@ def get_detection_dataset_dicts(
             `min_keypoints`. Set to 0 to do nothing.
         proposal_files (list[str]): if given, a list of object proposal files
             that match each dataset in `dataset_names`.
+
+    Returns:
+        list[dict]: a list of dicts following the standard dataset dict format.
     """
     assert len(dataset_names)
     dataset_dicts = [DatasetCatalog.get(dataset_name) for dataset_name in dataset_names]
@@ -228,10 +231,8 @@ def get_detection_dataset_dicts(
     dataset_dicts = list(itertools.chain.from_iterable(dataset_dicts))
 
     has_instances = "annotations" in dataset_dicts[0]
-    # Keep images without instance-level GT if the dataset has semantic labels.
-    if filter_empty and has_instances and "sem_seg_file_name" not in dataset_dicts[0]:
+    if filter_empty and has_instances:
         dataset_dicts = filter_images_with_only_crowd_annotations(dataset_dicts)
-
     if min_keypoints > 0 and has_instances:
         dataset_dicts = filter_images_with_few_keypoints(dataset_dicts, min_keypoints)