- Descrizione :
Set di dati Omniglot per l'apprendimento one-shot. Questo set di dati contiene 1623 caratteri manoscritti diversi, provenienti da 50 alfabeti diversi.
Documentazione aggiuntiva : Esplora su Papers With Code
Pagina iniziale : https://github.com/brendenlake/omniglot/
Codice sorgente :
tfds.image_classification.OmniglotVersioni :
-
3.0.0(predefinito): Nuova API divisa ( https://tensorflow.org/datasets/splits )
-
Dimensione download :
17.95 MiBDimensione del set di dati :
12.29 MiBMemorizzazione automatica nella cache ( documentazione ): Sì
Divisioni :
| Diviso | Esempi |
|---|---|
'small1' | 2.720 |
'small2' | 3.120 |
'test' | 13.180 |
'train' | 19.280 |
- Struttura delle caratteristiche :
FeaturesDict({
'alphabet': ClassLabel(shape=(), dtype=int64, num_classes=50),
'alphabet_char_id': int64,
'image': Image(shape=(105, 105, 3), dtype=uint8),
'label': ClassLabel(shape=(), dtype=int64, num_classes=1623),
})
- Documentazione delle funzionalità :
| Caratteristica | Classe | Forma | Tipo D | Descrizione |
|---|---|---|---|---|
| CaratteristicheDict | ||||
| alfabeto | Etichetta di classe | int64 | ||
| alfabeto_carattere_id | Tensore | int64 | ||
| immagine | Immagine | (105, 105, 3) | uint8 | |
| etichetta | Etichetta di classe | int64 |
Chiavi supervisionate (vedere il documento
as_supervised):('image', 'label')Figura ( tfds.show_examples ):

- Esempi ( tfds.as_dataframe ):
- Citazione :
@article{lake2015human,
title={Human-level concept learning through probabilistic program induction},
author={Lake, Brenden M and Salakhutdinov, Ruslan and Tenenbaum, Joshua B},
journal={Science},
volume={350},
number={6266},
pages={1332--1338},
year={2015},
publisher={American Association for the Advancement of Science}
}