車196

Cars データセットには、196 クラスの自動車の 16,185 枚の画像が含まれています。データは 8,144 のトレーニング イメージと 8,041 のテスト イメージに分割され、各クラスは大まかに 50 対 50 に分割されています。クラスは通常、メーカー、モデル、年式のレベルです (例: 2012 テスラ モデル S または 2012 BMW M3 クーペ)。

スプリット
'test' 8,041
'train' 8,144
  • 機能構造:
FeaturesDict({
    'bbox': BBoxFeature(shape=(4,), dtype=float32),
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=196),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
bbox BBoxFeature (4) float32
ID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{KrauseStarkDengFei-Fei_3DRR2013,
title = {3D Object Representations for Fine-Grained Categorization},
booktitle = {4th International IEEE Workshop on  3D Representation and Recognition (3dRR-13)},
year = {2013},
address = {Sydney, Australia},
author = {Jonathan Krause and Michael Stark and Jia Deng and Li Fei-Fei}
}