def __getitem__(self, ind):
if isinstance(self.indexes, int):
if ind != 0:
raise OutOfBoundsError(f"Got index {ind} for dataset of length 1")
index = self.indexes
else:
index = self.indexes[ind]
self._init_ds()
d = {}
for key in self._ds._tensors.keys():
if key not in self.key_list:
continue
split_key = key.split("/")
cur = d
for i in range(1, len(split_key) - 1):
if split_key[i] not in cur.keys():
cur[split_key[i]] = {}
cur = cur[split_key[i]]
item = self._get_active_item(key, index)
if not isinstance(item, bytes) and not isinstance(item, str):
t = item
if self.inplace:
if t.dtype == "uint16":
t = t.astype("int32")
elif t.dtype == "uint32" or t.dtype == "uint64":
t = t.astype("int64")
t = torch.tensor(t)
cur[split_key[-1]] = t
d = self._do_transform(d)
if self.inplace & (self.output_type != dict) & (isinstance(d, dict)):
d = self.output_type(d.values())
return d
def create(cls, obj, context=None, bounded=False):
typ = type(obj)
if typ is Bbox:
obj = obj
elif typ in (list, tuple):
obj = Bbox.from_slices(obj, context, bounded)
elif typ is Vec:
obj = Bbox.from_vec(obj)
elif typ is str:
obj = Bbox.from_filename(obj)
elif typ is dict:
obj = Bbox.from_dict(obj)
else:
raise NotImplementedError(
"{} is not a Bbox convertible type.".format(typ))
if context and bounded:
if not context.contains_bbox(obj):
raise OutOfBoundsError(
"{} did not fully contain the specified bounding box {}.".
format(context, obj))
return obj
def check_bounds(val, low, high):
if val > high or val < low:
raise OutOfBoundsError(
'Value {} cannot be outside of inclusive range {} to {}'.format(
val, low, high))
return val