イメージネット_a

  • 説明

ImageNet-A は、新しいデータを収集し、ResNet-50 モデルが正しく分類できなかった画像のみを保持することによって取得された、ImageNet ラベルでラベル付けされた画像のセットです。詳細については、論文を参照してください。

ラベル空間はImageNet2012と同じです。各例は、次のキーを持つ辞書として表されます。

スプリット
'test' 7,500
  • 機能の構造:
FeaturesDict({
   
'file_name': Text(shape=(), dtype=string),
   
'image': Image(shape=(None, None, 3), dtype=uint8),
   
'label': ClassLabel(shape=(), dtype=int64, num_classes=1000),
})
  • 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
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
@article{hendrycks2019nae,
  title
={Natural Adversarial Examples},
  author
={Dan Hendrycks and Kevin Zhao and Steven Basart and Jacob Steinhardt and Dawn Song},
  journal
={arXiv preprint arXiv:1907.07174},
  year
={2019}
}