ビジュアライゼーション: Know Your Data で探索する
説明:
ワンショット学習用の Omniglot データセット。このデータセットには、50 の異なるアルファベットからの 1623 の異なる手書き文字が含まれています。
追加のドキュメント:コード を使用したペーパーの探索
ソース コード:
tfds.image_classification.Omniglot
バージョン:
-
3.0.0
(デフォルト): 新しい分割 API ( https://tensorflow.org/datasets/splits )
-
ダウンロードサイズ:
17.95 MiB
データセットのサイズ:
12.29 MiB
自動キャッシュ(ドキュメント): はい
スプリット:
スプリット | 例 |
---|---|
'small1' | 2,720 |
'small2' | 3,120 |
'test' | 13,180 |
'train' | 19,280 |
- 機能構造:
FeaturesDict({
'alphabet': ClassLabel(shape=(), dtype=int64, num_classes=50),
'alphabet_char_id': int64,
'image': Image(shape=(105, 105, 3), dtype=uint8),
'label': ClassLabel(shape=(), dtype=int64, num_classes=1623),
})
- 機能のドキュメント:
特徴 | クラス | 形 | Dtype | 説明 |
---|---|---|---|---|
特徴辞書 | ||||
アルファベット | クラスラベル | int64 | ||
alphabet_char_id | テンソル | int64 | ||
画像 | 画像 | (105、105、3) | uint8 | |
ラベル | クラスラベル | int64 |
監視されたキー(
as_supervised
docを参照):('image', 'label')
図( tfds.show_examples ):
- 例( tfds.as_dataframe ):
- 引用:
@article{lake2015human,
title={Human-level concept learning through probabilistic program induction},
author={Lake, Brenden M and Salakhutdinov, Ruslan and Tenenbaum, Joshua B},
journal={Science},
volume={350},
number={6266},
pages={1332--1338},
year={2015},
publisher={American Association for the Advancement of Science}
}