THULab/LSY-lab_libero_plus_spatial_texture_all_ur5e_failures

LIBERO Plus Spatial Texture All UR5e Failures TsFile

This dataset is an Apache TsFile conversion of LSY-lab/libero_plus_spatial_texture_all_ur5e_failures, a LeRobot v3.0 UR5e manipulation dataset containing failure trajectories across ten spatial-texture variants of a LIBERO-Plus bowl-placement task. The source repository is published by LSY-lab; the original uploader/author is Maximilian Christof (max-chr).

Source Dataset and Attribution

• Original dataset: LSY-lab/libero_plus_spatial_texture_all_ur5e_failures
• Source organization: LSY-lab
• Original uploader/author: Maximilian Christof (max-chr)
• License: Apache-2.0
• Robot type: ur5e; LeRobot codebase: v3.0
• Sampling rate: 20 fps; split: train
• The source card does not provide a dataset-specific paper citation or BibTeX entry.
• Source frame layout: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• Source video layout: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

Videos and Frame Alignment

Videos are not copied to this TsFile repository. Retrieve them from the original videos/ tree. Align numeric rows to MP4 frames using episode_index and frame_index; meta/episodes/chunk-000/file-000.parquet contains each stream's chunk/file and timestamp ranges.

Minimal Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/lsy_lab_libero_plus_spatial_texture_all_ur5e_failures.tsfile")
table_name = "lsy_lab_libero_plus_spatial_texture_all_ur5e_failures"
columns = [
    "episode_index", "task_index", "frame_index", "sample_index",
    "action_0", "observation_state_0",
]
with reader.query_table(table_name, columns, batch_size=65536) as result:
    batch = result.read_arrow_batch()
    print(batch.to_pandas().head())
reader.close()

The dataset-specific YAML and converter remain available locally under D:/code/config/ and D:/code/scripts/; validation reports are local artifacts and are not part of the TsFile upload set.