uiuc_d3필드

  • 설명 :

사무실 책상, 식기 등 정리

나뉘다
'train' 192
  • 기능 구조 :
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'file_path': Text(shape=(), dtype=string),
    }),
    'steps': Dataset({
        'action': Tensor(shape=(3,), dtype=float32),
        'discount': Scalar(shape=(), dtype=float32),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'language_embedding': Tensor(shape=(512,), dtype=float32),
        'language_instruction': Text(shape=(), dtype=string),
        'observation': FeaturesDict({
            'depth_1': Image(shape=(360, 640, 1), dtype=uint16),
            'depth_2': Image(shape=(360, 640, 1), dtype=uint16),
            'depth_3': Image(shape=(360, 640, 1), dtype=uint16),
            'depth_4': Image(shape=(360, 640, 1), dtype=uint16),
            'image_1': Image(shape=(360, 640, 3), dtype=uint8),
            'image_2': Image(shape=(360, 640, 3), dtype=uint8),
            'image_3': Image(shape=(360, 640, 3), dtype=uint8),
            'image_4': Image(shape=(360, 640, 3), dtype=uint8),
            'state': Tensor(shape=(4, 4), dtype=float32),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • 기능 문서 :
특징 수업 모양 Dtype 설명
특징Dict
에피소드_메타데이터 특징Dict
에피소드_메타데이터/파일_경로 텍스트 원본 데이터 파일의 경로입니다.
단계 데이터세트
단계/작업 텐서 (삼,) float32 마지막 프레임에서 로봇 변위
걸음수/할인 스칼라 float32 할인이 제공되면 기본값은 1입니다.
걸음수/is_first 텐서 부울
걸음수/is_last 텐서 부울
단계/is_terminal 텐서 부울
단계/언어_임베딩 텐서 (512,) float32 코나 언어 임베딩. https://tfhub.dev/google/universal-sentence-encoder-large/5를 참조하세요.
단계/언어_지시 텍스트 언어 교육.
단계/관찰 특징Dict
단계/관찰/깊이_1 영상 (360, 640, 1) 단위16 카메라 1 깊이 관찰.
단계/관찰/깊이_2 영상 (360, 640, 1) 단위16 카메라 2 깊이 관찰.
단계/관찰/깊이_3 영상 (360, 640, 1) 단위16 카메라 3 깊이 관찰.
단계/관찰/깊이_4 영상 (360, 640, 1) 단위16 카메라 4 깊이 관찰.
단계/관찰/image_1 영상 (360, 640, 3) uint8 카메라 1 RGB 관찰.
단계/관찰/image_2 영상 (360, 640, 3) uint8 카메라 2 RGB 관찰.
단계/관찰/image_3 영상 (360, 640, 3) uint8 카메라 3 RGB 관찰.
단계/관찰/image_4 영상 (360, 640, 3) uint8 카메라 4 RGB 관찰.
단계/관찰/상태 텐서 (4, 4) float32 로봇 엔드 이펙터 상태
걸음수/보상 스칼라 float32 제공되는 경우 보상, 데모의 마지막 단계에서 1개.
@article{wang2023d3field,
  title={D^3Field: Dynamic 3D Descriptor Fields for Generalizable Robotic Manipulation},
  author={Wang, Yixuan and Li, Zhuoran and Zhang, Mingtong and Driggs-Campbell, Katherine and Wu, Jiajun and Fei-Fei, Li and Li, Yunzhu},
  journal={arXiv preprint arXiv:},
  year={2023},
}