def test_empty(self):
x = torch.zeros(4, 1, 34, 34)
to_dense = ME.MinkowskiToDenseTensor(x.shape)
# Convert to sparse data
sparse_data = ME.to_sparse(x)
dense_data = to_dense(sparse_data)
self.assertEqual(dense_data.shape, x.shape)
def __init__(self, config={}, **kwargs):
self.config = self.__class__.default_config()
self.config.update(config)
self.config.update(kwargs)
if self.config.data_root is None:
self.n_images = 0
if self.config.img_size is not None:
self.img_size = self.config.img_size
self.coords = []
self.feats = []
self.n_channels = 1
else:
raise ValueError(
"If data_root not given, the img_size must be specified in the config"
)
self.labels = torch.zeros((0, 1), dtype=torch.long)
else:
# load HDF5 Mnist3d dataset
dataset_filepath = os.path.join(self.config.data_root, '3Dmnist',
'full_dataset_vectors.h5')
with h5py.File(dataset_filepath, 'r') as file:
if self.config.split == "train":
X = file["X_train"][:]
Y = file["y_train"][:]
elif self.config.split in ["valid", "test"]:
X = file["X_test"][:]
Y = file["y_test"][:]
self.n_images = int(X.shape[0])
self.has_labels = True
self.labels = torch.LongTensor(Y)
self.img_size = (16, 16, 16)
self.n_channels = 1
self.coords = []
self.feats = []
images = torch.Tensor(X).float().reshape((
-1,
self.n_channels,
) + self.img_size)
images_sparse = ME.to_sparse(images)
for idx in range(len(images)):
self.coords.append(images_sparse.coordinates_at(idx))
self.feats.append(images_sparse.features_at(idx))
if self.config.preprocess is not None:
self.coords, self.feats = self.config.preprocess(
self.coords, self.feats)
# data augmentation boolean
self.data_augmentation = self.config.data_augmentation
if self.data_augmentation:
self.augment = []
# the user can additionally specify a transform in the config
self.transform = self.config.transform
self.target_transform = self.config.target_transform
