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 80, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
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.
Multiview RGB clips and camera calibration
100 selected clips, each 15 seconds at 15 fps (225 frames), with 47 valid camera views at 2048 × 1536. The recommended download format is one HEVC archive per clip, under hevc/C001.tar through hevc/C100.tar.
Upload is in progress. Consult dataset_index.json (available: true) or hevc_upload_status.json for verified, available clips. The index is updated after each archive has been committed and its size and SHA-256 verified remotely. A just-committed archive may appear before its index update.
Original JPEG archives already uploaded under data/ are retained unchanged. Further JPEG uploads have stopped. Each index row has jpeg_archive_path when that original archive exists; null means it is unavailable. HEVC uploads use a separate namespace and do not replace the originals. The index now describes the preferred HEVC format.
Download one clip
from huggingface_hub import hf_hub_download
archive = hf_hub_download(
repo_id="Sulwon/anonymous_dataset_lih_orh_0920",
filename="hevc/C001.tar",
repo_type="dataset",
local_dir="downloads",
)
print(archive)
mkdir -p hevc_clips
tar -xf downloads/hevc/C001.tar -C hevc_clips
Archives are independent and contain:
C001/
videos/<camera_serial>.mp4 # 47 videos
video_manifest.json # camera and per-frame correspondence
cameras.json
cam_param/intrinsics.json
cam_param/extrinsics.json
timestamps/<camera_serial>.json
Extract JPEG and HEVC archives into separate directories to preserve their format-specific camera path metadata. JPEG archives contain manifest.json and rgb/<camera_serial>/<frame_index>.jpg.
Video and camera conventions
The existing HEVC videos are packaged without re-encoding: NVENC HEVC, CQ 18, preset p4, HQ tuning, VBR, GOP 15 and 2 B-frames. This is lossy compression of the extracted JPEG frames. There is no resizing, cropping or undistortion. The source video payload totals about 156 GB; tar and camera metadata add a small overhead. Tar archives are not compressed again.
Use encoded_frame_index (0–224) to index decoded MP4 frames. video_frame_index refers to the original source recording and must not be used as the extracted MP4 index. The manifest preserves raw frame identifiers, frame IDs and capture timestamps. The nominal clip duration is 15 seconds; the first-to-last sampled time span is 224/15 seconds.
RGB images remain distorted. Use K_original and the provided distortion model and coefficients for projection into the supplied frames. K_undistort is for a separately undistorted image. Camera matrices include T_cam_from_world and T_world_from_camera; follow the coordinate and unit conventions recorded in each manifest. Numeric source intrinsics and extrinsics are preserved byte-for-byte.
There are 48 calibrated cameras, but only 47 valid video streams. Camera 23012639 has no valid source frames: its calibration is retained with valid: false, and no synthetic video is provided. All archive paths are relative to the clip directory. No depth, masks, reconstructions, model results or processing scripts are included.
Integrity and provenance
Every packaged member is checked against the prepared SHA-256 inventory. Intrinsics/extrinsics are also checked against the manifest calibration hashes. Each completed archive has archive_bytes and archive_sha256 in the index, verified against the Hub file metadata. Videos were previously checked during encoding for 225 frames, 15 fps, original dimensions and successful decoding; packaging does not re-run video encoding.
Clip source identifiers and selected frame windows remain in the index and manifests. Licensing and citation details have not yet been supplied by the dataset owner; public availability alone does not specify a reuse license.
- Downloads last month
- 170