ビッグアースネット

  • 説明

BigEarthNet は、新しい大規模な Sentinel-2 ベンチマーク アーカイブであり、590,326 個の Sentinel-2 イメージ パッチで構成されています。地上の画像パッチのサイズは 1.2 x 1.2 km で、チャネル解像度に応じて画像サイズが変わります。これは、43 個の不均衡なラベルを含むマルチラベル データセットです。

BigEarthNet を構築するために、ヨーロッパの 10 か国 (オーストリア、ベルギー、フィンランド、アイルランド、コソボ、リトアニア、ルクセンブルク、ポルトガル、セルビア、スイス) で 2017 年 6 月から 2018 年 5 月の間に取得された 125 個の Sentinel-2 タイルが最初に選択されました。すべてのタイルは、Sentinel-2 レベル 2A プロダクト生成およびフォーマット ツール (sen2cor) によって大気補正されました。次に、それらは 590,326 個の重複しない画像パッチに分割されました。各画像パッチには、2018 年の CORINE 土地被覆データベース (CLC 2018) から提供された複数の土地被覆クラス (つまり、マルチラベル) によって注釈が付けられました。

バンドとピクセル解像度 (メートル単位):

  • B01: 沿岸エアロゾル。 60m
  • B02: ブルー; 10メートル
  • B03: グリーン; 10メートル
  • B04: レッド; 10メートル
  • B05: 植生のレッドエッジ。 20メートル
  • B06: 植生のレッドエッジ。 20メートル
  • B07: 植生のレッドエッジ。 20メートル
  • B08:近赤外; 10メートル
  • B09: 水蒸気。 60m
  • B11: SWIR; 20m
  • B12: SWIR; 20メートル
  • B8A: 狭い近赤外。 20メートル

ライセンス: コミュニティ データ使用許諾契約 - 許容的、バージョン 1.0。

URL: http://bigearth.net/

スプリット
'train' 590,326
  • 引用
@article{Sumbul2019BigEarthNetAL,
  title={BigEarthNet: A Large-Scale Benchmark Archive For Remote Sensing Image Understanding},
  author={Gencer Sumbul and Marcela Charfuelan and Beg{"u}m Demir and Volker Markl},
  journal={CoRR},
  year={2019},
  volume={abs/1902.06148}
}

bigearthnet/rgb (デフォルト設定)

  • 構成の説明: Sentinel-2 RGB チャネル

  • データセットのサイズ: 14.07 GiB

  • 機能の構造:

FeaturesDict({
    'filename': Text(shape=(), dtype=string),
    'image': Image(shape=(120, 120, 3), dtype=uint8),
    'labels': Sequence(ClassLabel(shape=(), dtype=int64, num_classes=43)),
    'metadata': FeaturesDict({
        'acquisition_date': Text(shape=(), dtype=string),
        'coordinates': FeaturesDict({
            'lrx': int64,
            'lry': int64,
            'ulx': int64,
            'uly': int64,
        }),
        'projection': Text(shape=(), dtype=string),
        'tile_source': Text(shape=(), dtype=string),
    }),
})
  • 機能ドキュメント:
特徴クラスDタイプ説明
特徴辞書
ファイル名文章
画像画像(120、120、3) uint8
ラベルシーケンス(クラスラベル) (なし、) int64
メタデータ特徴辞書
メタデータ/取得日文章
メタデータ/座標特徴辞書
メタデータ/座標/lrxテンソルint64
メタデータ/座標/lryテンソルint64
メタデータ/座標/ウルクステンソルint64
メタデータ/座標/ulyテンソルint64
メタデータ/投影文章
メタデータ/タイルソース文章

視覚化

ビッグアースネット/すべて

  • 構成の説明: 13 Sentinel-2 チャネル

  • データセットのサイズ: 176.63 GiB

  • 機能の構造:

FeaturesDict({
    'B01': Tensor(shape=(20, 20), dtype=float32),
    'B02': Tensor(shape=(120, 120), dtype=float32),
    'B03': Tensor(shape=(120, 120), dtype=float32),
    'B04': Tensor(shape=(120, 120), dtype=float32),
    'B05': Tensor(shape=(60, 60), dtype=float32),
    'B06': Tensor(shape=(60, 60), dtype=float32),
    'B07': Tensor(shape=(60, 60), dtype=float32),
    'B08': Tensor(shape=(120, 120), dtype=float32),
    'B09': Tensor(shape=(20, 20), dtype=float32),
    'B11': Tensor(shape=(60, 60), dtype=float32),
    'B12': Tensor(shape=(60, 60), dtype=float32),
    'B8A': Tensor(shape=(60, 60), dtype=float32),
    'filename': Text(shape=(), dtype=string),
    'labels': Sequence(ClassLabel(shape=(), dtype=int64, num_classes=43)),
    'metadata': FeaturesDict({
        'acquisition_date': Text(shape=(), dtype=string),
        'coordinates': FeaturesDict({
            'lrx': int64,
            'lry': int64,
            'ulx': int64,
            'uly': int64,
        }),
        'projection': Text(shape=(), dtype=string),
        'tile_source': Text(shape=(), dtype=string),
    }),
})
  • 機能ドキュメント:
特徴クラスDタイプ説明
特徴辞書
B01テンソル(20、20) float32
B02テンソル(120、120) float32
B03テンソル(120、120) float32
B04テンソル(120、120) float32
B05テンソル(60、60) float32
B06テンソル(60、60) float32
B07テンソル(60、60) float32
B08テンソル(120、120) float32
B09テンソル(20、20) float32
B11テンソル(60、60) float32
B12テンソル(60、60) float32
B8Aテンソル(60、60) float32
ファイル名文章
ラベルシーケンス(クラスラベル) (なし、) int64
メタデータ特徴辞書
メタデータ/取得日文章
メタデータ/座標特徴辞書
メタデータ/座標/lrxテンソルint64
メタデータ/座標/lryテンソルint64
メタデータ/座標/ウルクステンソルint64
メタデータ/座標/ulyテンソルint64
メタデータ/投影文章
メタデータ/タイルソース文章