deep1b、deep1b

  • 説明:

コサイン距離を使用した近似最近傍検索の事前トレーニング済み埋め込み。このデータセットは、次の 2 つの分割で構成されています。

  1. 'database': 9,990,000 個のデータ ポイントで構成され、それぞれに次の機能があります: 'embedding' (96 float)、'index' (int64)、'neighbors' (空のリスト)。
  2. 'test': 10,000 個のデータ ポイントで構成され、それぞれに次の機能があります。 )
スプリット
'database' 9,990,000
'test' 10,000
  • 機能構造:
FeaturesDict({
    'embedding': Tensor(shape=(96,), dtype=float32),
    'index': Scalar(shape=(), dtype=int64),
    'neighbors': Sequence({
        'distance': Scalar(shape=(), dtype=float32),
        'index': Scalar(shape=(), dtype=int64),
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
埋め込みテンソル(96) float32
索引スカラーint64分割内のインデックス。
隣人順序テスト分割でのみ使用できる、計算された隣接ノード。
隣人/距離スカラーfloat32近隣距離。
ネイバー/インデックススカラーint64ネイバー インデックス。
  • 引用
@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}
}