cmu_franka_exploration_dataset_converted_externally_to_rlds

  • 설명 :

장난감 부엌을 탐험하는 프랑카

나뉘다
'train' 199
  • 기능 구조 :
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'file_path': Text(shape=(), dtype=string),
    }),
    'steps': Dataset({
        'action': Tensor(shape=(8,), 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({
            'highres_image': Image(shape=(480, 640, 3), dtype=uint8),
            'image': Image(shape=(64, 64, 3), dtype=uint8),
        }),
        'reward': Scalar(shape=(), dtype=float32),
        'structured_action': Tensor(shape=(8,), 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 (8,) float32 로봇 액션은 [엔드 이펙터 위치3x, 엔드 이펙터 방향3x, 그리퍼 액션1x, 에피소드 종료1x]로 구성됩니다.
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/highres_image 영상 (480, 640, 3) uint8 고해상도 메인 카메라 관찰
단계/관찰/이미지 영상 (64, 64, 3) uint8 메인 카메라 RGB 관찰.
걸음수/보상 스칼라 float32 제공되는 경우 보상, 데모의 마지막 단계에서 1개.
단계/구조화된_작업 텐서 (8,) float32 휴먼 비디오의 구조화된 세계 모델에 설명된 하이브리드 어포던스 및 엔드 이펙터 제어로 구성된 구조화된 동작입니다.
  • 인용 :
@inproceedings{mendonca2023structured,
              title={Structured World Models from Human Videos},
              author={Mendonca, Russell  and Bahl, Shikhar and Pathak, Deepak},
              journal={RSS},
              year={2023}
            }