oxford_iiit_pet

Oxford-IIIT ペット データセットは、37 カテゴリのペット画像データセットで、各クラスに約 200 枚の画像があります。画像には、スケール、ポーズ、および照明に大きなバリエーションがあります。すべての画像には、品種のグラウンド トゥルース アノテーションが関連付けられています。

スプリット
'test' 3,669
'train' 3,680
  • 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=37),
    'segmentation_mask': Image(shape=(None, None, 1), dtype=uint8),
    'species': ClassLabel(shape=(), dtype=int64, num_classes=2),
})
  • 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
セグメンテーション_マスク画像(なし、なし、1) uint8
種族クラスラベルint64
  • 引用
@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",
}