cifar10

Dataset CIFAR-10 terdiri dari 60000 citra berwarna 32x32 dalam 10 kelas, dengan 6000 citra per kelas. Ada 50.000 gambar pelatihan dan 10.000 gambar uji.

Membelah Contoh
'test' 10.000
'train' 50.000
  • Struktur fitur :
FeaturesDict({
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(32, 32, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=10),
})
  • Dokumentasi fitur :
Fitur Kelas Membentuk Dtype Keterangan
fiturDict
Indo Teks rangkaian
gambar Gambar (32, 32, 3) uint8
label LabelKelas int64

Visualisasi

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