move_mnist

  • 설명 :

손으로 쓴 숫자의 MNIST 데이터베이스 이동 변형. 이는 모델 성능을 보고하기 위해 작성자가 사용하는 데이터입니다. MNIST 데이터 세트에서 훈련/검증 데이터를 생성 tfds.video.moving_mnist.image_as_moving_sequence 를 참조하세요.

나뉘다
'test' 10,000
  • 기능 구조 :
FeaturesDict({
    'image_sequence': Video(Image(shape=(64, 64, 1), dtype=uint8)),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
이미지 시퀀스 동영상(이미지) (20, 64, 64, 1) uint8
  • 인용 :
@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}
}