- opis :
Ruchomy wariant bazy danych MNIST z odręcznymi cyframi. Są to dane używane przez autorów do raportowania wydajności modelu. Zobacz tfds.video.moving_mnist.image_as_moving_sequence aby uzyskać informacje na temat generowania danych treningowych/walidacyjnych z zestawu danych MNIST.
Dodatkowa dokumentacja : Przeglądaj dokumenty z kodem na
Strona główna : http://www.cs.toronto.edu/~nitish/unsupervised_video/
Kod źródłowy :
tfds.video.MovingMnistWersje :
-
1.0.0(domyślnie): nowy podzielony interfejs API ( https://tensorflow.org/datasets/splits )
-
Rozmiar pliku do pobrania :
781.25 MiBRozmiar zestawu danych :
91.70 MiBAutomatyczne buforowanie ( dokumentacja ): Tak
Podziały :
| Rozdzielać | Przykłady |
|---|---|
'test' | 10 000 |
- Struktura funkcji :
FeaturesDict({
'image_sequence': Video(Image(shape=(64, 64, 1), dtype=uint8)),
})
- Dokumentacja funkcji :
| Funkcja | Klasa | Kształt | Typ D | Opis |
|---|---|---|---|---|
| FunkcjeDict | ||||
| sekwencja_obrazów | Wideo (obraz) | (20, 64, 64, 1) | uint8 |
Klucze nadzorowane (Zobacz dokument
as_supervised):NoneRysunek ( tfds.show_examples ): Nieobsługiwany.
Przykłady ( tfds.as_dataframe ):
- Cytat :
@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}
}