Carsデータセットには、196クラスの車の16,185枚の画像が含まれています。データは8,144のトレーニング画像と8,041のテスト画像に分割され、各クラスは大まかに50〜50の分割に分割されています。クラスは通常、メーカー、モデル、年のレベルで、たとえば2012テスラモデルSまたは2012BMWM3クーペです。
ホームページ: https ://ai.stanford.edu/~jkrause/cars/car_dataset.html
バージョン:
-
2.0.0
:初期リリース 2.0.1
:ウェブサイトのURLの更新2.1.0
(デフォルト):バグの修正https://github.com/tensorflow/datasets/issues/3927
-
ダウンロードサイズ:
1.82 GiB
データセットサイズ:
1.85 GiB
自動キャッシュ(ドキュメント):いいえ
分割:
スプリット | 例 |
---|---|
'test' | 8,041 |
'train' | 8,144 |
- 素性構造:
FeaturesDict({
'bbox': BBoxFeature(shape=(4,), dtype=tf.float32),
'id': Text(shape=(), dtype=tf.string),
'image': Image(shape=(None, None, 3), dtype=tf.uint8),
'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=196),
})
- 機能ドキュメント:
特徴 | クラス | 形 | Dtype | 説明 |
---|---|---|---|---|
FeaturesDict | ||||
bbox | BBoxFeature | (4、) | tf.float32 | |
id | 文章 | tf.string | ||
画像 | 画像 | (なし、なし、3) | tf.uint8 | |
ラベル | ClassLabel | tf.int64 |
監視対象キー(
as_supervised
docを参照):(('image', 'label')
図( tfds.show_examples ):
- 例( tfds.as_dataframe ):
- 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}
}