옴니글로트

원샷 학습을 위한 Omniglot 데이터 세트. 이 데이터 세트에는 50개의 서로 다른 알파벳에서 1623개의 서로 다른 필기 문자가 포함되어 있습니다.

나뉘다
'small1' 2,720
'small2' 3,120
'test' 13,180
'train' 19,280
  • 기능 구조 :
FeaturesDict({
    'alphabet': ClassLabel(shape=(), dtype=int64, num_classes=50),
    'alphabet_char_id': int64,
    'image': Image(shape=(105, 105, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=1623),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
알파벳 클래스 레이블 int64
alphabet_char_id 텐서 int64
영상 영상 (105, 105, 3) uint8
상표 클래스 레이블 int64

심상

  • 인용 :
@article{lake2015human,
  title={Human-level concept learning through probabilistic program induction},
  author={Lake, Brenden M and Salakhutdinov, Ruslan and Tenenbaum, Joshua B},
  journal={Science},
  volume={350},
  number={6266},
  pages={1332--1338},
  year={2015},
  publisher={American Association for the Advancement of Science}
}