- 説明:
DomainNet データセットは、写真 (実物)、絵画、クリップアート、クイックドロー、インフォグラフ、スケッチなど、6 つの異なるドメインからの画像で構成されています。ドメインごとに 48K ~ 172K の画像 (合計 600K) があり、345 のクラスに分類されています。
DomainNet のこの TFDS バージョンでは、クリーン バージョンが使用されます。
追加のドキュメント:コードを使用したペーパーの探索
ホームページ: http://ai.bu.edu/DomainNet/
バージョン:
-
1.0.0(デフォルト): 初期リリース。
-
自動キャッシュ(ドキュメント): いいえ
機能構造:
FeaturesDict({
'image': Image(shape=(None, None, 3), dtype=uint8),
'label': ClassLabel(shape=(), dtype=int64, num_classes=345),
})
- 機能のドキュメント:
| 特徴 | クラス | 形 | Dtype | 説明 |
|---|---|---|---|---|
| 特徴辞書 | ||||
| 画像 | 画像 | (なし、なし、3) | uint8 | |
| ラベル | クラスラベル | int64 |
監視されたキー(
as_superviseddocを参照):('image', 'label')引用:
@inproceedings{peng2019moment,
title={Moment matching for multi-source domain adaptation},
author={Peng, Xingchao and Bai, Qinxun and Xia, Xide and Huang, Zijun and Saenko, Kate and Wang, Bo},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={1406--1415},
year={2019}
}
domainnet/real (デフォルト構成)
ダウンロードサイズ:
5.65 GiBデータセットサイズ:
5.75 GiBスプリット:
| スプリット | 例 |
|---|---|
'test' | 52,041 |
'train' | 120,906 |
- 図( tfds.show_examples ):

- 例( tfds.as_dataframe ):
ドメインネット/絵画
ダウンロードサイズ:
3.43 GiBデータセットサイズ:
3.30 GiBスプリット:
| スプリット | 例 |
|---|---|
'test' | 21,850 |
'train' | 50,416 |
- 図( tfds.show_examples ):

- 例( tfds.as_dataframe ):
ドメインネット/クリップアート
ダウンロードサイズ:
1.19 GiBデータセットサイズ:
1.27 GiBスプリット:
| スプリット | 例 |
|---|---|
'test' | 14,604 |
'train' | 33,525 |
- 図( tfds.show_examples ):

- 例( tfds.as_dataframe ):
ドメインネット/クイックドロー
ダウンロードサイズ:
445.65 MiBデータセットのサイズ:
435.16 MiBスプリット:
| スプリット | 例 |
|---|---|
'test' | 51,750 |
'train' | 120,750 |
- 図( tfds.show_examples ):

- 例( tfds.as_dataframe ):
ドメインネット/インフォグラフ
ダウンロードサイズ:
4.04 GiBデータセットサイズ:
4.17 GiBスプリット:
| スプリット | 例 |
|---|---|
'test' | 15,582 |
'train' | 36,023 |
- 図( tfds.show_examples ):

- 例( tfds.as_dataframe ):
ドメインネット/スケッチ
ダウンロードサイズ:
2.46 GiBデータセットサイズ:
2.55 GiBスプリット:
| スプリット | 例 |
|---|---|
'test' | 20,916 |
'train' | 48,212 |
- 図( tfds.show_examples ):

- 例( tfds.as_dataframe ):