オックスフォード_iiit_pet

  • 説明

Oxford-IIIT ペット データセットは、クラスごとに約 200 枚の画像を含む 37 カテゴリのペット画像データセットです。画像にはスケール、ポーズ、照明が大きく異なります。すべての画像には、品種と種に関するグラウンド トゥルースの注釈が関連付けられています。さらに、トレーニング分割用に頭部境界ボックスが提供され、このデータセットを単純な物体検出タスクに使用できるようになります。テスト分割では、境界ボックスは空です。

スプリット
'test' 3,669
'train' 3,680
  • 機能の構造:
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),
})
  • 機能ドキュメント:
特徴クラスDタイプ説明
特徴辞書
ファイル名文章
head_bbox BBox機能(4,) float32
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
クラスラベルint64
@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",
}