khang123452/manipulation-init-frame-descriptions
Manipulation init-frame / description pairs 1000 (init frame image, task description) pairs randomly sampled (seed=42) from the manipulation task family of nvidia/PhysicalAI-WorldModel-Synthetic-Embodied-Robot-Scenes. Each row is the first frame of a simulated manipulation clip paired with its task description text, drawn from the following generators within the manipulation task family: DreamZero, MimicGen (AgiBot G1 / Fourier GR-1 / Galbot G1), and Simulario. Columns:… See the full description on the dataset page: https://huggingface.co/datasets/khang123452/manipulation-init-frame-descriptions.
Manipulation init-frame / description pairs
1000 (init frame image, task description) pairs randomly sampled (seed=42) from the manipulation task family of nvidia/PhysicalAI-WorldModel-Synthetic-Embodied-Robot-Scenes.
Each row is the first frame of a simulated manipulation clip paired with its task description text, drawn from the following generators within the manipulation task family: DreamZero, MimicGen (AgiBot G1 / Fourier GR-1 / Galbot G1), and Simulario.
Columns:
file_name: image file (first frame of the source clip)text: task descriptiondesc_field: which source metadata fieldtextwas taken fromtask_family,source_key,source_shard: provenance back to the source dataset
