시파르10

CIFAR-10 데이터 세트는 10개 클래스의 60000개 32x32 컬러 이미지로 구성되며 클래스당 6000개의 이미지가 있습니다. 50000개의 훈련 이미지와 10000개의 테스트 이미지가 있습니다.

나뉘다
'test' 10,000
'train' 50,000
  • 기능 구조 :
FeaturesDict({
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(32, 32, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=10),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
ID 텍스트
영상 영상 (32, 32, 3) uint8
상표 클래스 레이블 int64

심상

  • 인용 :
@TECHREPORT{Krizhevsky09learningmultiple,
    author = {Alex Krizhevsky},
    title = {Learning multiple layers of features from tiny images},
    institution = {},
    year = {2009}
}