def __init__(self):
self.images, features = _load_small_norb_chunks(
SMALLNORB_TEMPLATE, SMALLNORB_CHUNKS)
self.image_leng = self.images.shape[0]
self.factor_sizes = [5, 10, 9, 18, 6]
# Instances are not part of the latent space.
self.latent_factor_indices = [0, 2, 3, 4]
self.num_total_factors = features.shape[1]
self.index = util.StateSpaceAtomIndex(self.factor_sizes, features)
self.state_space = util.SplitDiscreteStateSpace(
self.factor_sizes, self.latent_factor_indices)
def __init__(self):
self.factor_sizes = [4, 24, 183]
features = cartesian([np.array(list(range(i))) for i in self.factor_sizes])
self.latent_factor_indices = [0, 1, 2]
self.num_total_factors = features.shape[1]
self.index = util.StateSpaceAtomIndex(self.factor_sizes, features)
self.factor_bases = self.index.factor_bases
self.state_space = util.SplitDiscreteStateSpace(self.factor_sizes,
self.latent_factor_indices)
self.data_shape = [64, 64, 3]
self.images = self._load_data()
def __init__(self, mode="mpi3d_toy"):
if mode == "mpi3d_toy":
mpi3d_path = os.path.join(
os.environ.get("DISENTANGLEMENT_LIB_DATA", "."), "mpi3d_toy",
"mpi3d_toy.npz")
if not tf.io.gfile.exists(mpi3d_path):
raise ValueError(
"Dataset '{}' not found. Make sure the dataset is publicly available and downloaded correctly."
.format(mode))
else:
with tf.io.gfile.GFile(mpi3d_path, "rb") as f:
data = np.load(f)
self.factor_sizes = [4, 4, 2, 3, 3, 40, 40]
elif mode == "mpi3d_realistic":
mpi3d_path = os.path.join(
os.environ.get("DISENTANGLEMENT_LIB_DATA", "."),
"mpi3d_realistic", "mpi3d_realistic.npz")
if not tf.io.gfile.exists(mpi3d_path):
raise ValueError(
"Dataset '{}' not found. Make sure the dataset is publicly available and downloaded correctly."
.format(mode))
else:
with tf.io.gfile.GFile(mpi3d_path, "rb") as f:
data = np.load(f)
self.factor_sizes = [4, 4, 2, 3, 3, 40, 40]
elif mode == "mpi3d_real":
mpi3d_path = os.path.join(
os.environ.get("DISENTANGLEMENT_LIB_DATA", "."), "mpi3d_real",
"mpi3d_real.npz")
if not tf.io.gfile.exists(mpi3d_path):
raise ValueError(
"Dataset '{}' not found. Make sure the dataset is publicly available and downloaded correctly."
.format(mode))
else:
with tf.io.gfile.GFile(mpi3d_path, "rb") as f:
data = np.load(f)
self.factor_sizes = [6, 6, 2, 3, 3, 40, 40]
else:
raise ValueError("Unknown mode provided.")
self.images = data["images"]
self.image_leng = self.images.shape[0]
self.latent_factor_indices = [0, 1, 2, 3, 4, 5, 6]
self.num_total_factors = 7
self.state_space = util.SplitDiscreteStateSpace(
self.factor_sizes, self.latent_factor_indices)
self.factor_bases = np.prod(self.factor_sizes) / np.cumprod(
self.factor_sizes)
def __init__(self):
data = h5py.File(SHAPES3D_PATH, 'r')
# images = data["images"].value
# labels = data["labels"].value
images = data['images'][()]
labels = data["labels"][()]
n_samples = 480000
self.images = (
images.reshape([n_samples, 64, 64, 3]).astype(np.float32) / 255.)
features = labels.reshape([n_samples, 6])
self.factor_sizes = [10, 10, 10, 8, 4, 15]
self.latent_factor_indices = list(range(6))
self.num_total_factors = features.shape[1]
self.state_space = util.SplitDiscreteStateSpace(
self.factor_sizes, self.latent_factor_indices)
self.factor_bases = np.prod(self.factor_sizes) / np.cumprod(
self.factor_sizes)
def __init__(self):
with tf.gfile.GFile(SHAPES3D_PATH, "rb") as f:
# Data was saved originally using python2, so we need to set the encoding.
data = np.load(f, encoding="latin1")
images = data["images"]
labels = data["labels"]
n_samples = np.prod(images.shape[0:6])
self.images = (
images.reshape([n_samples, 64, 64, 3]).astype(np.float32) / 255.)
self.image_leng = self.images.shape[0]
features = labels.reshape([n_samples, 6])
self.factor_sizes = [10, 10, 10, 8, 4, 15]
self.latent_factor_indices = list(range(6))
self.num_total_factors = features.shape[1]
self.state_space = util.SplitDiscreteStateSpace(self.factor_sizes,
self.latent_factor_indices)
self.factor_bases = np.prod(self.factor_sizes) / np.cumprod(
self.factor_sizes)
def __init__(self, latent_factor_indices=None):
# By default, all factors (including shape) are considered ground truth
# factors.
if latent_factor_indices is None:
latent_factor_indices = list(range(6))
self.latent_factor_indices = latent_factor_indices
self.data_shape = [64, 64, 1]
# Load the data so that we can sample from it.
with gfile.Open(DSPRITES_PATH, "rb") as data_file:
# Data was saved originally using python2, so we need to set the encoding.
data = np.load(data_file, encoding="latin1", allow_pickle=True)
self.images = np.array(data["imgs"])
self.factor_sizes = np.array(data["metadata"][()]["latents_sizes"],
dtype=np.int64)
self.full_factor_sizes = [1, 3, 6, 40, 32, 32]
self.factor_bases = np.prod(self.factor_sizes) / np.cumprod(
self.factor_sizes)
self.state_space = util.SplitDiscreteStateSpace(
self.factor_sizes, self.latent_factor_indices)