cifar10

Il set di dati CIFAR-10 è costituito da 60000 immagini a colori 32x32 in 10 classi, con 6000 immagini per classe. Sono disponibili 50000 immagini di addestramento e 10000 immagini di test.

Diviso Esempi
'test' 10.000
'train' 50.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(32, 32, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=10),
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
id Testo corda
Immagine Immagine (32, 32, 3) uint8
etichetta ClassLabel int64

Visualizzazione

  • Citazione :
@TECHREPORT{Krizhevsky09learningmultiple,
    author = {Alex Krizhevsky},
    title = {Learning multiple layers of features from tiny images},
    institution = {},
    year = {2009}
}