場所いっぱい
コレクションでコンテンツを整理
必要に応じて、コンテンツの保存と分類を行います。
Places データセットは、人間の視覚認識の原則に従って設計されています。私たちの目標は、シーン コンテキスト、オブジェクト認識、アクションとイベントの予測、理論による推論など、高レベルの視覚的理解タスクのために人工システムをトレーニングするために使用できる視覚的知識のコアを構築することです。
場所のセマンティック カテゴリは、その機能によって定義されます。ラベルは、環境のエントリ レベルを表します。説明すると、データセットには寝室や通りなどのさまざまなカテゴリがあります。これは、家の寝室、ホテルの寝室、または子供部屋で同じように行動したり、次に何が起こるかについて同じ予測を行ったりしないためです。 Places には、400 以上のユニークなシーン カテゴリからなる 1,000 万枚を超える画像が含まれています。このデータセットには、クラスごとに 5000 から 30,000 のトレーニング画像が含まれており、実際の発生頻度と一致しています。 Places データセットは、畳み込みニューラル ネットワーク (CNN) を使用して、シーン中心のベンチマークで新しい最先端のパフォーマンスを確立することを目標に、さまざまなシーン認識タスクのディープ シーン機能の学習を可能にします。
ここでは、学術研究および教育目的で Places Database とトレーニング済み CNN を提供しています。
FeaturesDict({
'filename': Text(shape=(), dtype=string),
'image': Image(shape=(256, 256, 3), dtype=uint8),
'label': ClassLabel(shape=(), dtype=int64, num_classes=435),
})
特徴 | クラス | 形 | Dtype | 説明 |
---|
| 特徴辞書 | | | |
ファイル名 | 文章 | | ストリング | |
画像 | 画像 | (256, 256, 3) | uint8 | |
ラベル | クラスラベル | | int64 | |

@article{zhou2017places,
title={Places: A 10 million Image Database for Scene Recognition},
author={Zhou, Bolei and Lapedriza, Agata and Khosla, Aditya and Oliva, Aude and Torralba, Antonio},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
year={2017},
publisher={IEEE}
}
特に記載のない限り、このページのコンテンツはクリエイティブ・コモンズの表示 4.0 ライセンスにより使用許諾されます。コードサンプルは Apache 2.0 ライセンスにより使用許諾されます。詳しくは、Google Developers サイトのポリシーをご覧ください。Java は Oracle および関連会社の登録商標です。
最終更新日 2022-12-16 UTC。
[null,null,["最終更新日 2022-12-16 UTC。"],[],[],null,["# placesfull\n\n\u003cbr /\u003e\n\n- **Description**:\n\nThe Places dataset is designed following principles of human visual cognition.\nOur goal is to build a core of visual knowledge that can be used to train\nartificial systems for high-level visual understanding tasks, such as scene\ncontext, object recognition, action and event prediction, and theory-of-mind\ninference.\n\nThe semantic categories of Places are defined by their function: the labels\nrepresent the entry-level of an environment. To illustrate, the dataset has\ndifferent categories of bedrooms, or streets, etc, as one does not act the same\nway, and does not make the same predictions of what can happen next, in a home\nbedroom, an hotel bedroom or a nursery. In total, Places contains more than 10\nmillion images comprising 400+ unique scene categories. The dataset features\n5000 to 30,000 training images per class, consistent with real-world frequencies\nof occurrence. Using convolutional neural networks (CNN), Places dataset allows\nlearning of deep scene features for various scene recognition tasks, with the\ngoal to establish new state-of-the-art performances on scene-centric benchmarks.\n\nHere we provide the Places Database and the trained CNNs for academic research\nand education purposes.\n\n- **Homepage** : \u003chttp://places2.csail.mit.edu/\u003e\n\n- **Source code** :\n [`tfds.datasets.placesfull.Builder`](https://github.com/tensorflow/datasets/tree/master/tensorflow_datasets/datasets/placesfull/placesfull_dataset_builder.py)\n\n- **Versions**:\n\n - **`1.0.0`** (default): No release notes.\n- **Download size** : `143.56 GiB`\n\n- **Dataset size** : `136.56 GiB`\n\n- **Auto-cached**\n ([documentation](https://www.tensorflow.org/datasets/performances#auto-caching)):\n No\n\n- **Splits**:\n\n| Split | Examples |\n|-----------|------------|\n| `'train'` | 10,653,087 |\n\n- **Feature structure**:\n\n FeaturesDict({\n 'filename': Text(shape=(), dtype=string),\n 'image': Image(shape=(256, 256, 3), dtype=uint8),\n 'label': ClassLabel(shape=(), dtype=int64, num_classes=435),\n })\n\n- **Feature documentation**:\n\n| Feature | Class | Shape | Dtype | Description |\n|----------|--------------|---------------|--------|-------------|\n| | FeaturesDict | | | |\n| filename | Text | | string | |\n| image | Image | (256, 256, 3) | uint8 | |\n| label | ClassLabel | | int64 | |\n\n- **Supervised keys** (See\n [`as_supervised` doc](https://www.tensorflow.org/datasets/api_docs/python/tfds/load#args)):\n `('image', 'label', 'filename')`\n\n- **Figure**\n ([tfds.show_examples](https://www.tensorflow.org/datasets/api_docs/python/tfds/visualization/show_examples)):\n\n- **Examples** ([tfds.as_dataframe](https://www.tensorflow.org/datasets/api_docs/python/tfds/as_dataframe)):\n\nDisplay examples... \n\n- **Citation**:\n\n @article{zhou2017places,\n title={Places: A 10 million Image Database for Scene Recognition},\n author={Zhou, Bolei and Lapedriza, Agata and Khosla, Aditya and Oliva, Aude and Torralba, Antonio},\n journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},\n year={2017},\n publisher={IEEE}\n }"]]