Update coco.py

add 'skip' configurations

coco.py

@@ -565,13 +565,14 @@ PANOPTIC_FEATURE = datasets.Features(
 class CocoConfig(datasets.BuilderConfig):
   """BuilderConfig for CocoConfig."""
 
-  def __init__(self, features, splits=None, has_panoptic=False, **kwargs):
+  def __init__(self, features, splits=None, has_panoptic=False, skip_empty_annotations=False, **kwargs):
     super(CocoConfig, self).__init__(
         **kwargs
     )
     self.features = features
     self.splits = splits
     self.has_panoptic = has_panoptic
+    self.skip_empty_annotations = skip_empty_annotations
 
 
 # Copied from https://github.com/tensorflow/datasets/blob/master/tensorflow_datasets/object_detection/coco.py
@@ -659,6 +660,55 @@ class Coco(datasets.GeneratorBasedBuilder):
               ),
           ],
       ),
+      CocoConfig(
+          name='2017_skip',
+          features=DETECTION_FEATURE,
+          description=_CONFIG_DESCRIPTION.format(year=2017),
+          version=_VERSION,
+          skip_empty_annotations=True,
+          splits=[
+              Split(
+                  name=datasets.Split.TRAIN,
+                  images='train2017',
+                  annotations='annotations_trainval2017',
+                  annotation_type=AnnotationType.BBOXES,
+              ),
+              Split(
+                  name=datasets.Split.VALIDATION,
+                  images='val2017',
+                  annotations='annotations_trainval2017',
+                  annotation_type=AnnotationType.BBOXES,
+              ),
+              Split(
+                  name=datasets.Split.TEST,
+                  images='test2017',
+                  annotations='image_info_test2017',
+                  annotation_type=AnnotationType.NONE,
+              ),
+          ],
+      ),
+      CocoConfig(
+          name='2017_panoptic_skip',
+          features=PANOPTIC_FEATURE,
+          description=_CONFIG_DESCRIPTION.format(year=2017),
+          version=_VERSION,
+          has_panoptic=True,
+          skip_empty_annotations=True,
+          splits=[
+              Split(
+                  name=datasets.Split.TRAIN,
+                  images='train2017',
+                  annotations='panoptic_annotations_trainval2017',
+                  annotation_type=AnnotationType.PANOPTIC,
+              ),
+              Split(
+                  name=datasets.Split.VALIDATION,
+                  images='val2017',
+                  annotations='panoptic_annotations_trainval2017',
+                  annotation_type=AnnotationType.PANOPTIC,
+              ),
+          ],
+      ),
   ]
 
     DEFAULT_CONFIG_NAME = "2017"
@@ -753,6 +803,8 @@ class Coco(datasets.GeneratorBasedBuilder):
         elif annotation_type == AnnotationType.NONE:  # No annotation for test sets
             instance_filename = 'image_info_{}.json'
 
+        skip_empty_annotations = self.config.skip_empty_annotations
+        
         # Load the annotations (label names, images metadata,...)
         instance_path = os.path.join(
             annotation_dir,
@@ -760,13 +812,6 @@ class Coco(datasets.GeneratorBasedBuilder):
             instance_filename.format(split_name),
         )
         coco_annotation = ANNOTATION_CLS[annotation_type](instance_path)
-        # Each category is a dict:
-        # {
-        #    'id': 51,  # From 1-91, some entry missing
-        #    'name': 'bowl',
-        #    'supercategory': 'kitchen',
-        # }
-        categories = coco_annotation.categories
         # Each image is a dict:
         # {
         #     'id': 262145,
@@ -835,6 +880,8 @@ class Coco(datasets.GeneratorBasedBuilder):
 
             if not instances:
                 annotation_skipped += 1
+                if skip_empty_annotations:
+                    continue
 
             def build_bbox(x, y, width, height):
                 # pylint: disable=cell-var-from-loop