oxford_iiit_pet

  • 설명 :

Oxford-IIIT 애완동물 데이터세트는 각 클래스에 대해 대략 200개의 이미지가 포함된 37개 카테고리의 애완동물 이미지 데이터세트입니다. 이미지의 크기, 포즈, 조명은 매우 다양합니다. 모든 이미지에는 품종 및 종에 대한 관련 실측 주석이 있습니다. 또한 훈련 분할을 위해 머리 경계 상자가 제공되므로 이 데이터 세트를 간단한 객체 감지 작업에 사용할 수 있습니다. 테스트 분할에서는 경계 상자가 비어 있습니다.

나뉘다
  • 기능 구조 :
FeaturesDict({
    'file_name': Text(shape=(), dtype=string),
    'head_bbox': BBoxFeature(shape=(4,), dtype=float32),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=37),
    'segmentation_mask': Image(shape=(None, None, 1), dtype=uint8),
    'species': ClassLabel(shape=(), dtype=int64, num_classes=2),
})
  • 기능 문서 :
특징 수업 모양 Dtype 설명
특징Dict
파일_이름 텍스트
head_bbox B박스특징 (4,) float32
영상 영상 (없음, 없음, 3) uint8
상표 클래스 라벨 정수64
분할_마스크 영상 (없음, 없음, 1) uint8
클래스 라벨 정수64
@InProceedings{parkhi12a,
  author       = "Parkhi, O. M. and Vedaldi, A. and Zisserman, A. and Jawahar, C.~V.",
  title        = "Cats and Dogs",
  booktitle    = "IEEE Conference on Computer Vision and Pattern Recognition",
  year         = "2012",
}