Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
image
imagewidth (px)
256
640
End of preview. Expand in Data Studio

LIBERO-Pro Rollout Videos

Public front-view rollout videos and preview images for LIBERO-Pro. This dataset consolidates the media used by the project website, including model evaluation rollouts, training demonstration previews, and the benchmark overview video.

Contents

  • assets/rollouts/models/: model, task, and perturbation video galleries.
  • assets/rollouts/training/: the existing training demonstration previews.
  • assets/promo/: the benchmark overview video and its poster.
  • model-rollouts.json: model gallery paths with the original task and perturbation coverage.
  • training-media.json: paths for the training previews selected by the website.
  • files.jsonl: file paths, byte counts, and SHA-256 checksums.

There are 7,240 MP4 videos. Existing galleries retain their original published bytes, resolution, and frame rate. Some earlier galleries contain lower-resolution sources, while later recordings are native 640 x 640. Both successful and unsuccessful evaluation rollouts are included. These visual examples do not replace the reported repeated-trial evaluation results.

The CloudRobo-WM gallery adds 339 submitted videos at 640 x 640, covering eight Base tasks and all 331 applicable task–perturbation combinations. For display, these videos retain only the left camera view: the right camera view and letterbox bars are removed. The original 320 x 320 view is resized to 640 x 640 with Lanczos interpolation. Frame rate, orientation and complete duration are preserved. H.264 encoding is lossless relative to the cropped and resized frames, verified over every decoded frame. Preview images show each video’s first frame.

Browse and download

The model galleries organize evaluation videos by model, task, and perturbation. Public perturbation identifiers are S01-S22 and D01-D20; the indexes retain the supported combinations for each gallery.

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="racingemperor/LIBERO-Pro-Videos",
    repo_type="dataset",
    allow_patterns=["assets/rollouts/models/fastwam/**", "model-rollouts.json"],
    local_dir="libero-pro-videos",
)

MP4 files can also be opened individually through their Hugging Face resolve URLs. Preview images use WebP.

Related resources

This repository provides rendered media. Robot demonstration HDF5 files and simulation evaluation assets remain in their respective datasets.

Downloads last month
702