robosuite_panda_pick_place_can

  • Sự miêu tả :

Những bộ dữ liệu này đã được tạo bằng môi trường PickPlaceCan của trình mô phỏng cánh tay robot robosuite . Các bộ dữ liệu về con người được ghi lại bởi một người vận hành bằng cách sử dụng RLDS Creator và bộ điều khiển gamepad.

Các bộ dữ liệu tổng hợp đã được ghi lại bằng thư viện EnvLogger .

Các bộ dữ liệu tuân theo định dạng RLDS để thể hiện các bước và các tập.

Các tập phim bao gồm 400 bước. Trong mỗi tập, một thẻ sẽ được thêm vào khi nhiệm vụ hoàn thành, thẻ này được lưu trữ như một phần của siêu dữ liệu bước tùy chỉnh.

Lưu ý rằng, do sự phụ thuộc vào EnvLogger, việc tạo tập dữ liệu này hiện chỉ được hỗ trợ trên môi trường Linux.

@misc{ramos2021rlds,
      title={RLDS: an Ecosystem to Generate, Share and Use Datasets in Reinforcement Learning},
      author={Sabela Ramos and Sertan Girgin and Léonard Hussenot and Damien Vincent and Hanna Yakubovich and Daniel Toyama and Anita Gergely and Piotr Stanczyk and Raphael Marinier and Jeremiah Harmsen and Olivier Pietquin and Nikola Momchev},
      year={2021},
      eprint={2111.02767},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

robosuite_panda_pick_place_can/human_dc29b40a (cấu hình mặc định)

  • Mô tả cấu hình : Tập dữ liệu do con người tạo (50 tập).

  • Trang chủ : https://github.com/google-research/rlds

  • Kích thước tải xuống : 96.67 MiB

  • Kích thước tập dữ liệu : 407.24 MiB

  • Chia tách :

Tách ra Ví dụ
'train' 50
  • Cấu trúc tính năng :
FeaturesDict({
    'agent_id': string,
    'episode_id': string,
    'episode_index': int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float64),
        'discount': float64,
        'image': Image(shape=(None, None, 3), dtype=uint8),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=float64),
            'Can_quat': Tensor(shape=(4,), dtype=float64),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'object-state': Tensor(shape=(14,), dtype=float64),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=float64),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=float64),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=float64),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=float64),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=float64),
        }),
        'reward': float64,
        'tag:placed': bool,
    }),
})
  • Tài liệu tính năng :
Tính năng Lớp học Hình dạng loại D Sự miêu tả
Tính năngDict
đại lý_id Tenxơ sợi dây
tập_id Tenxơ sợi dây
tập_index Tenxơ int32
bước Tập dữ liệu
bước/hành động Tenxơ (7,) phao64
bước/giảm giá Tenxơ phao64
bước/hình ảnh Hình ảnh (Không, Không, 3) uint8
bước/is_first Tenxơ bool
bước/is_last Tenxơ bool
bước/is_terminal Tenxơ bool
bước/quan sát Tính năngDict
bước/quan sát/Can_pos Tenxơ (3,) phao64
bước/quan sát/Can_quat Tenxơ (4,) phao64
bước/quan sát/Can_to_robot0_eef_pos Tenxơ (3,) phao64
bước/quan sát/Can_to_robot0_eef_quat Tenxơ (4,) phao32
bước/quan sát/trạng thái đối tượng Tenxơ (14,) phao64
bước/quan sát/robot0_eef_pos Tenxơ (3,) phao64
bước/quan sát/robot0_eef_quat Tenxơ (4,) phao64
bước/quan sát/robot0_gripper_qpos Tenxơ (2,) phao64
bước/quan sát/robot0_gripper_qvel Tenxơ (2,) phao64
bước/quan sát/robot0_joint_pos_cos Tenxơ (7,) phao64
bước/quan sát/robot0_joint_pos_sin Tenxơ (7,) phao64
bước/quan sát/robot0_joint_vel Tenxơ (7,) phao64
bước/quan sát/robot0_propio-state Tenxơ (32,) phao64
bước/phần thưởng Tenxơ phao64
bước/thẻ:đặt Tenxơ bool

robosuite_panda_pick_place_can/human_images_dc29b40a

  • Mô tả cấu hình : Tập dữ liệu do con người tạo ra, bao gồm các hình ảnh với các góc camera khác nhau trong quá trình quan sát. Lưu ý rằng có thể mất một chút thời gian để tạo.

  • Trang chủ : https://www.tensorflow.org/datasets/catalog/robosuite_panda_pick_place_can

  • Kích thước tải xuống : 10.95 GiB

  • Kích thước tập dữ liệu : 7.53 GiB

  • Chia tách :

Tách ra Ví dụ
'train' 50
  • Cấu trúc tính năng :
FeaturesDict({
    'agent_id': string,
    'episode_id': string,
    'episode_index': int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float64),
        'discount': float64,
        'image': Image(shape=(None, None, 3), dtype=uint8),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=float64),
            'Can_quat': Tensor(shape=(4,), dtype=float64),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'agentview_image': Image(shape=(256, 256, 3), dtype=uint8),
            'birdview_image': Image(shape=(256, 256, 3), dtype=uint8),
            'object-state': Tensor(shape=(14,), dtype=float64),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=float64),
            'robot0_eye_in_hand_image': Image(shape=(256, 256, 3), dtype=uint8),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=float64),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=float64),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=float64),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=float64),
            'robot0_robotview_image': Image(shape=(256, 256, 3), dtype=uint8),
        }),
        'reward': float64,
        'tag:placed': bool,
    }),
})
  • Tài liệu tính năng :
Tính năng Lớp học Hình dạng loại D Sự miêu tả
Tính năngDict
đại lý_id Tenxơ sợi dây
tập_id Tenxơ sợi dây
tập_index Tenxơ int32
bước Tập dữ liệu
bước/hành động Tenxơ (7,) phao64
bước/giảm giá Tenxơ phao64
bước/hình ảnh Hình ảnh (Không, Không, 3) uint8
bước/is_first Tenxơ bool
bước/is_last Tenxơ bool
bước/is_terminal Tenxơ bool
bước/quan sát Tính năngDict
bước/quan sát/Can_pos Tenxơ (3,) phao64
bước/quan sát/Can_quat Tenxơ (4,) phao64
bước/quan sát/Can_to_robot0_eef_pos Tenxơ (3,) phao64
bước/quan sát/Can_to_robot0_eef_quat Tenxơ (4,) phao32
bước/quan sát/agentview_image Hình ảnh (256, 256, 3) uint8
bước/quan sát/birdview_image Hình ảnh (256, 256, 3) uint8
bước/quan sát/trạng thái đối tượng Tenxơ (14,) phao64
bước/quan sát/robot0_eef_pos Tenxơ (3,) phao64
bước/quan sát/robot0_eef_quat Tenxơ (4,) phao64
bước/quan sát/robot0_eye_in_hand_image Hình ảnh (256, 256, 3) uint8
bước/quan sát/robot0_gripper_qpos Tenxơ (2,) phao64
bước/quan sát/robot0_gripper_qvel Tenxơ (2,) phao64
bước/quan sát/robot0_joint_pos_cos Tenxơ (7,) phao64
bước/quan sát/robot0_joint_pos_sin Tenxơ (7,) phao64
bước/quan sát/robot0_joint_vel Tenxơ (7,) phao64
bước/quan sát/robot0_propio-state Tenxơ (32,) phao64
bước/quan sát/robot0_robotview_image Hình ảnh (256, 256, 3) uint8
bước/phần thưởng Tenxơ phao64
bước/thẻ:đặt Tenxơ bool

robosuite_panda_pick_place_can/synthetic_stochastic_sac_afe13968

  • Mô tả cấu hình : Tập dữ liệu tổng hợp được tạo bởi tác nhân ngẫu nhiên được đào tạo bằng SAC (200 tập).

  • Trang chủ : https://github.com/google-research/rlds

  • Kích thước tải xuống : 144.44 MiB

  • Kích thước tập dữ liệu : 622.86 MiB

  • Chia tách :

Tách ra Ví dụ
'train' 200
  • Cấu trúc tính năng :
FeaturesDict({
    'agent_id': string,
    'episode_id': string,
    'episode_index': int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float32),
        'discount': float64,
        'image': Image(shape=(None, None, 3), dtype=uint8),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=float32),
            'Can_quat': Tensor(shape=(4,), dtype=float32),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=float32),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'object-state': Tensor(shape=(14,), dtype=float32),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=float32),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=float32),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=float32),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=float32),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=float32),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=float32),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=float32),
        }),
        'reward': float64,
        'tag:placed': bool,
    }),
})
  • Tài liệu tính năng :
Tính năng Lớp học Hình dạng loại D Sự miêu tả
Tính năngDict
đại lý_id Tenxơ sợi dây
tập_id Tenxơ sợi dây
tập_index Tenxơ int32
bước Tập dữ liệu
bước/hành động Tenxơ (7,) phao32
bước/giảm giá Tenxơ phao64
bước/hình ảnh Hình ảnh (Không, Không, 3) uint8
bước/is_first Tenxơ bool
bước/is_last Tenxơ bool
bước/is_terminal Tenxơ bool
bước/quan sát Tính năngDict
bước/quan sát/Can_pos Tenxơ (3,) phao32
bước/quan sát/Can_quat Tenxơ (4,) phao32
bước/quan sát/Can_to_robot0_eef_pos Tenxơ (3,) phao32
bước/quan sát/Can_to_robot0_eef_quat Tenxơ (4,) phao32
bước/quan sát/trạng thái đối tượng Tenxơ (14,) phao32
bước/quan sát/robot0_eef_pos Tenxơ (3,) phao32
bước/quan sát/robot0_eef_quat Tenxơ (4,) phao32
bước/quan sát/robot0_gripper_qpos Tenxơ (2,) phao32
bước/quan sát/robot0_gripper_qvel Tenxơ (2,) phao32
bước/quan sát/robot0_joint_pos_cos Tenxơ (7,) phao32
bước/quan sát/robot0_joint_pos_sin Tenxơ (7,) phao32
bước/quan sát/robot0_joint_vel Tenxơ (7,) phao32
bước/quan sát/robot0_propio-state Tenxơ (32,) phao32
bước/phần thưởng Tenxơ phao64
bước/thẻ:đặt Tenxơ bool