bair_robot_pushing_small

  • 설명 :

이 데이터 세트에는 하나의 훈련 세트(기차)와 이전에 본(testseen) 객체와 보이지 않는(testnovel) 객체의 두 테스트 세트를 포함하여 대략 44,000개의 로봇 밀기 동작 예제가 포함되어 있습니다. 이것은 작은 64x64 버전입니다.

나뉘다
'test' 256
'train' 43,264
  • 기능 구조 :
Sequence({
    'action': Tensor(shape=(4,), dtype=float32),
    'endeffector_pos': Tensor(shape=(3,), dtype=float32),
    'image_aux1': Image(shape=(64, 64, 3), dtype=uint8),
    'image_main': Image(shape=(64, 64, 3), dtype=uint8),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
순서
동작 텐서 (4,) float32
endeffector_pos 텐서 (삼,) float32
이미지_aux1 영상 (64, 64, 3) uint8
image_main 영상 (64, 64, 3) uint8
  • 인용 :
@misc{1710.05268,
  Author = {Frederik Ebert and Chelsea Finn and Alex X. Lee and Sergey Levine},
  Title = {Self-Supervised Visual Planning with Temporal Skip Connections},
  Year = {2017},
  Eprint = {arXiv:1710.05268},
}