def __iter__(self):
for batch in take_by(self._extractor, self._batch_size):
inputs = np.array([item.image.data for item in batch])
inference = self._launcher.launch(inputs)
for item, annotations in zip(batch, inference):
yield self.wrap_item(item, annotations=annotations)
def transform_item(self, item):
if not item.has_image:
raise DatumaroError("Item %s: image info is required for this "
"transform" % (item.id, ))
h, w = item.image.size
xscale = self._width / float(w)
yscale = self._height / float(h)
new_size = (self._height, self._width)
resized_image = None
if item.image.has_data:
resized_image = self._lazy_resize_image(item.image, new_size)
resized_annotations = []
for ann in item.annotations:
if isinstance(ann, Bbox):
resized_annotations.append(
ann.wrap(
x=ann.x * xscale,
y=ann.y * yscale,
w=ann.w * xscale,
h=ann.h * yscale,
))
elif isinstance(ann, (Polygon, Points, PolyLine)):
resized_annotations.append(
ann.wrap(points=[
p for t in ((x * xscale, y * yscale)
for x, y in take_by(ann.points, 2))
for p in t
]))
elif isinstance(ann, Mask):
rescaled_mask = self._lazy_resize_mask(ann, new_size)
resized_annotations.append(ann.wrap(image=rescaled_mask))
elif isinstance(ann, (Caption, Label)):
resized_annotations.append(ann)
else:
assert False, f"Unexpected annotation type {type(ann)}"
return self.wrap_item(item,
image=resized_image,
annotations=resized_annotations)
def _load_annotations(self, ann, image_info=None, parsed_annotations=None):
if parsed_annotations is None:
parsed_annotations = []
ann_id = ann['id']
attributes = ann.get('attributes', {})
if 'score' in ann:
attributes['score'] = ann['score']
group = ann_id # make sure all tasks' annotations are merged
if self._task is CocoTask.instances or \
self._task is CocoTask.person_keypoints or \
self._task is CocoTask.stuff:
label_id = self._get_label_id(ann)
attributes['is_crowd'] = bool(ann['iscrowd'])
if self._task is CocoTask.person_keypoints:
keypoints = ann['keypoints']
points = []
visibility = []
for x, y, v in take_by(keypoints, 3):
points.append(x)
points.append(y)
visibility.append(v)
parsed_annotations.append(
Points(points,
visibility,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
segmentation = ann['segmentation']
if segmentation and segmentation != [[]]:
rle = None
if isinstance(segmentation, list):
if not self._merge_instance_polygons:
# polygon - a single object can consist of multiple parts
for polygon_points in segmentation:
parsed_annotations.append(
Polygon(points=polygon_points,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
else:
# merge all parts into a single mask RLE
rle = self._lazy_merged_mask(segmentation,
image_info['height'],
image_info['width'])
elif isinstance(segmentation['counts'], list):
# uncompressed RLE
img_h = image_info['height']
img_w = image_info['width']
mask_h, mask_w = segmentation['size']
if img_h == mask_h and img_w == mask_w:
rle = self._lazy_merged_mask([segmentation], mask_h,
mask_w)
else:
log.warning(
"item #%s: mask #%s "
"does not match image size: %s vs. %s. "
"Skipping this annotation.", image_info['id'],
ann_id, (mask_h, mask_w), (img_h, img_w))
else:
# compressed RLE
rle = segmentation
if rle:
parsed_annotations.append(
RleMask(rle=rle,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
else:
x, y, w, h = ann['bbox']
parsed_annotations.append(
Bbox(x,
y,
w,
h,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
elif self._task is CocoTask.labels:
label_id = self._get_label_id(ann)
parsed_annotations.append(
Label(label=label_id,
id=ann_id,
attributes=attributes,
group=group))
elif self._task is CocoTask.captions:
caption = ann['caption']
parsed_annotations.append(
Caption(caption, id=ann_id, attributes=attributes,
group=group))
else:
raise NotImplementedError()
return parsed_annotations