Datasetpublic

qqyang/sim-pen

ShadowHandPen RGB + raw rigid-contact tactile trajectories This dataset contains 5,000 successful ShadowHandPen PPO rollouts. Each episode stores synchronized RGB frames, robot hand joint/object state in trajectory_env0.npz, and dense EgoTouch-layout pressure in pressure_grids.npz. The pressure grids are computed from pairwise rigid contacts and Gaussian projection onto 217 valid taxels per hand. The tar files preserve the original collection shards. Shard-level duplicate RGB… See the full description on the dataset page: https://huggingface.co/datasets/qqyang/sim-pen.

sourceHugging Faceotherupdated 28d agoView on Hugging Face

0likes335downloads

rootREADME.md

README.md22 linesDownload Raw Back to root

1---2license: other3task_categories:4- reinforcement-learning5tags:6- dexterous-manipulation7- tactile8- rgb9- isaac-gym10---11 12# ShadowHandPen RGB + raw rigid-contact tactile trajectories13 14This dataset contains 5,000 successful `ShadowHandPen` PPO rollouts. Each15episode stores synchronized RGB frames, robot hand joint/object state in16`trajectory_env0.npz`, and dense EgoTouch-layout pressure in17`pressure_grids.npz`. The pressure grids are computed from pairwise rigid18contacts and Gaussian projection onto 217 valid taxels per hand.19 20The tar files preserve the original collection shards. Shard-level duplicate21RGB frame folders and QA-only videos are intentionally omitted.22