def bdc2dsmd_det_2d(annot_dir,
image_dir=None,
class2label=None,
ignore_label_name=True,
replace_ext=lambda x: x):
# N.B. annotation file name and image file name should be the same
num_classes = len(class2label) if class2label is not None else 1
filenames = mv.listdir(annot_dir)
empty_bboxes = np.zeros((0, 4), dtype=np.float32)
dsmd = {
replace_ext(filename): [empty_bboxes] * num_classes
for filename in filenames
}
for filename in filenames:
annot_filepath = mv.joinpath(annot_dir, filename)
bboxes = load_bdc_dr_bbox(
annot_filepath, lambda x: 0 if ignore_label_name else class2label)
for label, bbox in bboxes:
bbox = np.array(bbox, dtype=np.float32).reshape(-1, 4)
if dsmd[replace_ext(filename)][label].shape[0] == 0:
dsmd[replace_ext(filename)][label] = bbox
else:
dsmd[replace_ext(filename)][label] = np.append(
dsmd[replace_ext(filename)][label], bbox, axis=0)
return mv.make_dsmd(dsmd)
def save_cls_dsmd(dsmd_path, data, auto_mkdirs=True):
if auto_mkdirs:
mv.mkdirs(mv.parentdir(dsmd_path))
dsmd = mv.make_dsmd(data)
with open(dsmd_path, 'w') as fd:
for key, value in dsmd.items():
if mv.isarrayinstance(value): # handle multi-label case
value = ','.join([str(entry) for entry in value])
line = '%s,%s\n' % (str(key), str(value))
fd.write(line)
def load_cls_dsmd(dsmd_path):
data = {}
with open(dsmd_path, 'r') as fd:
for line in fd:
key, value = line.strip().split(',', 1)
try: # try to interpret annotation as int or list[int].
value = ast.literal_eval(value.strip())
except (SyntaxError, ValueError):
pass
data[key] = value
return mv.make_dsmd(data)
def save_det_dsmd(dsmd_path, data, class2label, auto_mkdirs=True):
""" Save dataset metadata to specified file.
Args:
dsmd_path (str): file path to save dataset metadata.
data (dict): dataset metadata, refer to 'load_dsmd'.
class2label (str or dict): class-to-label file or class2label dict.
auto_mkdirs (bool): If the parent folder of `file_path` does not
exist, whether to create it automatically.
"""
if auto_mkdirs:
mv.mkdirs(mv.parentdir(dsmd_path))
# get label->class mapping
if isinstance(class2label, str):
class2label = mv.load_c2l(class2label)
label2class = {value: key for key, value in class2label.items()}
# write dataset metadata loop
dsmd = mv.make_dsmd(data)
with open(dsmd_path, 'w') as fd:
for key, value in dsmd.items():
_write_record(fd, key, value, label2class)
def load_det_dsmd(dsmd_path, class2label):
""" load detection dataset metadata.
Args:
dsmd_path (str): dataset metadata file path.
class2label (str or dict): class-to-label file.
Return:
(OrderedDict): Loaded dsmd is a OrderedDict looks like
{
data/1.png: [
[bboxes of category 'cat' in ndarray of shape (n, 4)],
[bboxes of category 'dog' in ndarray of shape (n, 4)],
...
]
data/2.png: [
...
]
...
}
"""
data = {}
if isinstance(class2label, str):
class2label = mv.load_c2l(class2label)
# label should start from 0
assert min(class2label.values()) == 0
with open(dsmd_path, 'r') as fd:
for line in fd:
_update_data(data, line, class2label)
# convert bboxes from list of lists to ndarray
data = _convert_bboxes_format(data)
return mv.make_dsmd(data)