- คำอธิบาย :
ชุดข้อมูล Omniglot สำหรับการเรียนรู้แบบครั้งเดียว ชุดข้อมูลนี้ประกอบด้วยตัวอักษรที่เขียนด้วยลายมือ 1,623 ตัว จากตัวอักษร 50 แบบที่แตกต่างกัน
เอกสารเพิ่มเติม : สำรวจเอกสารด้วยรหัส
หน้าแรก : https://github.com/brendenlake/omniglot/
รหัสที่มา :
tfds.image_classification.Omniglotเวอร์ชัน :
-
3.0.0(ค่าเริ่มต้น): API การแยกใหม่ ( https://tensorflow.org/datasets/splits )
-
ขนาดดาวน์โหลด :
17.95 MiBขนาดชุดข้อมูล :
12.29 MiBแคชอัตโนมัติ ( เอกสารประกอบ ): ใช่
สปลิต :
| แยก | ตัวอย่าง |
|---|---|
'small1' | 2,720 |
'small2' | 3,120 |
'test' | 13,180 |
'train' | 19,280 |
- โครงสร้างคุณสมบัติ :
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),
})
- เอกสารประกอบคุณลักษณะ :
| คุณสมบัติ | ระดับ | รูปร่าง | ประเภท D | คำอธิบาย |
|---|---|---|---|---|
| ฟีเจอร์ดิกต์ | ||||
| ตัวอักษร | คลาสเลเบล | int64 | ||
| รหัสอักขระตัวอักษร | เทนเซอร์ | int64 | ||
| ภาพ | ภาพ | (105, 105, 3) | uint8 | |
| ฉลาก | คลาสเลเบล | int64 |
คีย์ที่ได้รับการดูแล (ดู เอกสาร
as_supervised):('image', 'label')รูปภาพ ( tfds.show_examples ):

- ตัวอย่าง ( tfds.as_dataframe ):
- 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}
}