GT-111/rlbench_v3_eef

RLBench-18 v3 EEF — Train Split (LeRobot v3.0)

RLBench 18-task PerAct dataset converted to LeRobot v3.0 with lossless H.264 video (CRF=0, yuv444p).

Format

• Per-task LeRobot v3.0 datasets — each task is a subdirectory with its own meta/, data/, videos/
• 8D EEF: xyz + quaternion (wxyz, scalar-first) + gripper
• State: absolute pose (wxyz passthrough from PerAct)
• Action: delta position + absolute quaternion (wxyz) + gripper
• 2 cameras: front view (observation.images.front_rgb) + wrist view (observation.images.wrist_rgb)
• 18 tasks, 100 episodes per task, ~374K total frames
• Raw values; use {task}/meta/norm_stats.json for per-task stats

Task List

close_jar, insert_onto_square_peg, light_bulb_in, meat_off_grill, open_drawer, place_cups, place_shape_in_shape_sorter, place_wine_at_rack_location, push_buttons, put_groceries_in_cupboard, put_item_in_drawer, put_money_in_safe, reach_and_drag, slide_block_to_color_target, stack_blocks, stack_cups, sweep_to_dustpan_of_size, turn_tap

Loading

from lerobot.datasets import LeRobotDataset

ds = LeRobotDataset("GT-111/rlbench_v3_eef/close_jar")
# Or load locally after download:
# ds = LeRobotDataset("GT-111/rlbench_v3_eef", revision="close_jar")

Acknowledgments

Data: PerAct format. Simulator: RLBench.

@inproceedings{shridhar2022peract,
  title={Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulation},
  author={Shridhar, Mohit and Manuelli, Lucas and Fox, Dieter},
  booktitle={CoRL},
  year={2022}
}
@article{james2019rlbench,
  title={RLBench: The Robot Learning Benchmark & Learning Environment},
  author={James, Stephen and Ma, Zicong and Rovick Arrojo, David and Davison, Andrew J},
  journal={IEEE Robotics and Automation Letters},
  year={2020}
}