moving_mnist

  • Mô tả :

Biến thể di chuyển của cơ sở dữ liệu MNIST của các chữ số viết tay. Đây là dữ liệu được các tác giả sử dụng để báo cáo hiệu suất mô hình. Xem tfds.video.moving_mnist.image_as_moving_sequence để tạo dữ liệu đào tạo/xác thực từ bộ dữ liệu MNIST.

Tách ra ví dụ
'test' 10.000
  • Cấu trúc tính năng :
FeaturesDict({
    'image_sequence': Video(Image(shape=(64, 64, 1), dtype=uint8)),
})
  • Tài liệu tính năng :
Tính năng Lớp Hình dạng Dtype Sự mô tả
Tính năngDict
image_sequence Video(Hình ảnh) (20, 64, 64, 1) uint8
  • trích dẫn :
@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}
}