Moving_mnist

  • 説明:

手書き数字の MNIST データベースの移動バリアント。これは、作成者がモデルのパフォーマンスを報告するために使用したデータです。 MNIST データセットからのトレーニング/検証データの生成については、 tfds.video.moving_mnist.image_as_moving_sequenceを参照してください。

スプリット
'test' 10,000
  • 機能構造:
FeaturesDict({
    'image_sequence': Video(Image(shape=(64, 64, 1), dtype=uint8)),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
image_sequence動画(画像) (20、64、64、1) uint8
  • 引用
@article{DBLP:journals/corr/SrivastavaMS15,
  author    = {Nitish Srivastava and
               Elman Mansimov and
               Ruslan Salakhutdinov},
  title     = {Unsupervised Learning of Video Representations using LSTMs},
  journal   = {CoRR},
  volume    = {abs/1502.04681},
  year      = {2015},
  url       = {http://arxiv.org/abs/1502.04681},
  archivePrefix = {arXiv},
  eprint    = {1502.04681},
  timestamp = {Mon, 13 Aug 2018 16:47:05 +0200},
  biburl    = {https://dblp.org/rec/bib/journals/corr/SrivastavaMS15},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}