omniglot

ชุดข้อมูล Omniglot สำหรับการเรียนรู้แบบ one-shot ชุดข้อมูลนี้มีอักขระที่เขียนด้วยลายมือที่แตกต่างกัน 1623 ตัวจาก 50 ตัวอักษรที่แตกต่างกัน

แยก ตัวอย่าง
'small1' 2,720
'small2' 3,120
'test' 13,180
'train' 19,280
  • โครงสร้างคุณลักษณะ :
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),
})
  • เอกสารคุณสมบัติ :
ลักษณะเฉพาะ ระดับ รูปร่าง Dประเภท คำอธิบาย
คุณสมบัติDict
ตัวอักษร ป้ายกำกับคลาส int64
ตัวอักษร_char_id เทนเซอร์ int64
ภาพ ภาพ (105, 105, 3) uint8
ฉลาก ป้ายกำกับคลาส int64

การสร้างภาพ

  • การอ้างอิง :
@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}
}