oxford_flowers102

  • 説明:

Oxford Flowers 102 データセットは、英国で一般的に見られる 102 の花のカテゴリから構成されています。各クラスは 40 ~ 258 枚の画像で構成されます。画像には大きな縮尺、ポーズ、光のバリエーションがあります。さらに、カテゴリ内で大きなバリエーションを持つカテゴリと、いくつかの非常に類似したカテゴリがあります。

データセットは、トレーニング セット、検証セット、およびテスト セットに分けられます。トレーニング セットと検証セットはそれぞれ、クラスごとに 10 個の画像で構成されます (それぞれ合計 1020 個の画像)。テスト セットは、残りの 6149 枚の画像 (クラスごとに最小 20 枚) で構成されます。

スプリット
'test' 6,149
'train' 1,020
'validation' 1,020
  • 機能構造:
FeaturesDict({
    'file_name': Text(shape=(), dtype=string),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=102),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
ファイル名文章ストリング
画像画像(なし、なし、3) uint8
ラベルクラスラベルint64

視覚化

  • 引用
@InProceedings{Nilsback08,
   author = "Nilsback, M-E. and Zisserman, A.",
   title = "Automated Flower Classification over a Large Number of Classes",
   booktitle = "Proceedings of the Indian Conference on Computer Vision, Graphics and Image Processing",
   year = "2008",
   month = "Dec"
}