cifar100

ชุดข้อมูลนี้เหมือนกับ CIFAR-10 ยกเว้นจะมี 100 คลาสที่มี 600 ภาพในแต่ละชุด มีภาพการฝึกอบรม 500 ภาพและภาพการทดสอบ 100 ภาพต่อชั้นเรียน 100 คลาสใน CIFAR-100 ถูกจัดกลุ่มเป็น 20 ซูเปอร์คลาส แต่ละภาพมาพร้อมกับป้ายกำกับ "ละเอียด" (คลาสที่ภาพนั้นอยู่) และป้ายกำกับ "หยาบ" (ซูเปอร์คลาสที่ภาพนั้นอยู่)

แยก ตัวอย่าง
'test' 10,000
'train' 50,000
  • โครงสร้างคุณลักษณะ :
FeaturesDict({
    'coarse_label': ClassLabel(shape=(), dtype=int64, num_classes=20),
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(32, 32, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=100),
})
  • เอกสารคุณสมบัติ :
ลักษณะเฉพาะ ระดับ รูปร่าง Dประเภท คำอธิบาย
คุณสมบัติDict
coarse_label ป้ายกำกับคลาส int64
รหัส ข้อความ สตริง
ภาพ ภาพ (32, 32, 3) uint8
ฉลาก ป้ายกำกับคลาส int64

การสร้างภาพ

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

ชุดข้อมูลนี้เหมือนกับ CIFAR-10 ยกเว้นจะมี 100 คลาสที่มี 600 ภาพในแต่ละชุด มีภาพการฝึกอบรม 500 ภาพและภาพการทดสอบ 100 ภาพต่อชั้นเรียน 100 คลาสใน CIFAR-100 ถูกจัดกลุ่มเป็น 20 ซูเปอร์คลาส แต่ละภาพมาพร้อมกับป้ายกำกับ "ละเอียด" (คลาสที่ภาพนั้นอยู่) และป้ายกำกับ "หยาบ" (ซูเปอร์คลาสที่ภาพนั้นอยู่)

แยก ตัวอย่าง
'test' 10,000
'train' 50,000
  • โครงสร้างคุณลักษณะ :
FeaturesDict({
    'coarse_label': ClassLabel(shape=(), dtype=int64, num_classes=20),
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(32, 32, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=100),
})
  • เอกสารคุณสมบัติ :
ลักษณะเฉพาะ ระดับ รูปร่าง Dประเภท คำอธิบาย
คุณสมบัติDict
coarse_label ป้ายกำกับคลาส int64
รหัส ข้อความ สตริง
ภาพ ภาพ (32, 32, 3) uint8
ฉลาก ป้ายกำกับคลาส int64

การสร้างภาพ

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