deep1b

  • 설명 :

코사인 거리를 사용하여 대략적인 최근접 검색을 위해 사전 훈련된 임베딩입니다. 이 데이터 세트는 두 가지 분할로 구성됩니다.

  1. '데이터베이스': 9,990,000개의 데이터 포인트로 구성되며 각 데이터 포인트에는 '임베딩'(96개 부동 소수점), '인덱스'(int64), '이웃'(빈 목록) 기능이 있습니다.
  2. '테스트': 10,000개의 데이터 포인트로 구성되며 각각은 '임베딩'(96개 부동 소수점), '인덱스'(int64), '이웃'(데이터베이스에서 가장 가까운 이웃의 '인덱스' 및 '거리' 목록)을 갖습니다. )
나뉘다
'database' 9,990,000
'test' 10,000
  • 기능 구조 :
FeaturesDict({
    'embedding': Tensor(shape=(96,), dtype=float32),
    'index': Scalar(shape=(), dtype=int64, description=Index within the split.),
    'neighbors': Sequence({
        'distance': Scalar(shape=(), dtype=float32, description=Neighbor distance.),
        'index': Scalar(shape=(), dtype=int64, description=Neighbor index.),
    }),
})
  • 기능 문서 :
특징 수업 모양 Dtype 설명
특징Dict
삽입 텐서 (96,) float32
색인 스칼라 정수64 분할 내의 인덱스입니다.
이웃 순서 테스트 분할에만 사용할 수 있는 계산된 이웃입니다.
이웃/거리 스칼라 float32 이웃 거리.
이웃/인덱스 스칼라 정수64 이웃 색인.
  • 인용 :
@inproceedings{babenko2016efficient,
  title={Efficient indexing of billion-scale datasets of deep descriptors},
  author={Babenko, Artem and Lempitsky, Victor},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={2055--2063},
  year={2016}
}