The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 78, in _split_generators
first_examples = list(islice(pipeline, self.NUM_EXAMPLES_FOR_FEATURES_INFERENCE))
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 54, in _get_pipeline_from_tar
current_example[field_name] = cls.DECODERS[data_extension](current_example[field_name])
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 316, in npy_loads
return numpy.lib.format.read_array(stream, allow_pickle=False)
~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/numpy/lib/_format_impl.py", line 820, in read_array
version = read_magic(fp)
File "/usr/local/lib/python3.14/site-packages/numpy/lib/_format_impl.py", line 245, in read_magic
raise ValueError(msg % (MAGIC_PREFIX, magic_str[:-2]))
ValueError: the magic string is not correct; expected b'\x93NUMPY', got b'd\xc2\x1f}\xda7'
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Art-Kubric
Art-Kubric contains 5,000 synthetic videos of articulated and rigid objects for point tracking and SE(3) motion estimation. Each video has 120 frames at 512×512 resolution and 60 fps, with 32,768 point tracks, per-link poses and segmentation maps.
100 shards, 50 scenes per shard. Download: 636.1 GB. Unpacked: approximately 2.9 TB.
Download
Download and unpack one shard:
pip install huggingface_hub numpy zstandard
git clone "https://github.com/mose3-tracker/Art-Kubric.git" art-kubric-code
hf download "mose3-tracker/Art-Kubric" scenes-0000.tar --repo-type dataset --local-dir art-kubric-data
mkdir -p art-kubric-data/scenes
tar -xf art-kubric-data/scenes-0000.tar -C art-kubric-data/scenes
python art-kubric-code/tools/unpack_release.py \
--src art-kubric-data/scenes --dst art-kubric-data/unpacked --all
For all shards, omit scenes-0000.tar from the download command and extract each tar.
index.json lists shard sizes, checksums and scene IDs.
Format
Each scene includes RGB frames, metric depth (depth_f32.npy), 2D/3D tracks and
visibility (*_trajs_*.npy, *_visibility.npy), link poses and segmentation
(se3/), and camera parameters (*_with_rank.npz).
3D tracks use world coordinates in metres; link poses are link-to-world transforms.
Camera extrinsics are world-to-camera in OpenGL axes (+X right, +Y up, -Z forward).
Flip their Y and Z rows for projection with shared_intrinsics. Match segmentation
labels to pose rows using link_actor_ids.npy and link_rp_seg_ids.npy.
License
Annotations: CC-BY-4.0. Upstream assets retain their original licenses, which users must also comply with. Asset geometry is not redistributed.
Citation
@inproceedings{cheng2026mose3,
title = {MoSE3: Learning World-Space SE(3) at Every Pixel},
author = {Cheng, Joanna Jiahuan and Li, Zhiyi and Xia, Tian and
Cai, Ruojin and Du, Yilun and Wang, Qianqian},
booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
year = {2026}
}
- Downloads last month
- -