- คำอธิบาย :
ImageNet-PI เป็นเวอร์ชันที่ติดป้ายกำกับใหม่ของชุดข้อมูล ImageNet มาตรฐาน ILSVRC2012 โดยป้ายกำกับนั้นได้มาจากเครือข่ายประสาทเทียมเชิงลึก 16 เครือข่ายที่มีสถาปัตยกรรมแตกต่างกัน ซึ่งได้รับการฝึกฝนล่วงหน้าบนชุดข้อมูล ILSVRC2012 มาตรฐาน โดยเฉพาะอย่างยิ่ง โมเดลที่ได้รับการฝึกฝนล่วงหน้าจะถูกดาวน์โหลดจาก tf.keras.applications
นอกเหนือจากป้ายกำกับใหม่แล้ว ImageNet-PI ยังให้ข้อมูลเมตาเกี่ยวกับกระบวนการติดป้ายกำกับในรูปแบบของความมั่นใจของแบบจำลองต่อป้ายกำกับ และข้อมูลเพิ่มเติมเกี่ยวกับแต่ละแบบจำลองอีกด้วย
สำหรับข้อมูลเพิ่มเติม โปรดดูที่: ImageNet-PI
หน้าหลัก : https://github.com/google-research-datasets/imagenet_pi/
โค้ดต้นฉบับ :
tfds.datasets.imagenet_pi.Builderเวอร์ชัน :
-
1.0.0(ค่าเริ่มต้น): เวอร์ชันเริ่มต้น
-
ขนาดไฟล์ดาวน์โหลด :
Unknown sizeขนาดชุดข้อมูล :
Unknown sizeคำแนะนำการดาวน์โหลดด้วยตนเอง : ชุดข้อมูลนี้ต้องการให้คุณดาวน์โหลดข้อมูลต้นฉบับด้วยตนเองไปยัง
download_config.manual_dir(ค่าเริ่มต้นคือ~/tensorflow_datasets/downloads/manual/):
โฟลเดอร์ manual_dir ควรมีไฟล์สองไฟล์ ได้แก่ ILSVRC2012_img_train.tar และ ILSVRC2012_img_val.tar คุณต้องลงทะเบียนที่ http://www.image-net.org/download-images เพื่อรับลิงก์สำหรับดาวน์โหลดชุดข้อมูลแคชอัตโนมัติ ( เอกสารประกอบ ): ไม่ทราบ
การแบ่งส่วน :
| แยก | ตัวอย่าง |
|---|
- โครงสร้างของฟีเจอร์ :
FeaturesDict({
'annotator_confidences': Tensor(shape=(16,), dtype=float32),
'annotator_labels': Tensor(shape=(16,), dtype=int64),
'clean_label': ClassLabel(shape=(), dtype=int64, num_classes=1000),
'file_name': Text(shape=(), dtype=string),
'image': Image(shape=(None, None, 3), dtype=uint8),
})
- เอกสารประกอบคุณสมบัติ :
| คุณสมบัติ | ระดับ | รูปร่าง | ประเภท D | คำอธิบาย |
|---|---|---|---|---|
| ฟีเจอร์ดิกต์ | ||||
| ความมั่นใจของผู้ให้คำอธิบาย | เทนเซอร์ | (16,) | ฟลอท32 | |
| ป้ายกำกับผู้ให้คำอธิบาย | เทนเซอร์ | (16,) | อินท์64 | |
| ฉลากสะอาด | ป้ายกำกับคลาส | อินท์64 | ||
| ชื่อไฟล์ | ข้อความ | สตริง | ||
| ภาพ | ภาพ | (ไม่มี, ไม่มี, 3) | uint8 |
คีย์ที่มีการกำกับดูแล (ดู เอกสาร
as_supervised):('image', 'annotator_labels')รูปภาพ ( tfds.show_examples ): ไม่รองรับ
ตัวอย่าง ( tfds.as_dataframe ): ขาดหายไป
อ้างอิง :
@inproceedings{tram,
author = {Mark Collier and
Rodolphe Jenatton and
Effrosyni Kokiopoulou and
Jesse Berent},
editor = {Kamalika Chaudhuri and
Stefanie Jegelka and
Le Song and
Csaba Szepesv{\'{a} }ri and
Gang Niu and
Sivan Sabato},
title = {Transfer and Marginalize: Explaining Away Label Noise with Privileged
Information},
booktitle = {International Conference on Machine Learning, {ICML} 2022, 17-23 July
2022, Baltimore, Maryland, {USA} },
series = {Proceedings of Machine Learning Research},
volume = {162},
pages = {4219--4237},
publisher = { {PMLR} },
year = {2022},
url = {https://proceedings.mlr.press/v162/collier22a.html},
}
@article{ILSVRC15,
Author = {Olga Russakovsky and Jia Deng and Hao Su and Jonathan Krause and Sanjeev Satheesh and Sean Ma and Zhiheng Huang and Andrej Karpathy and Aditya Khosla and Michael Bernstein and Alexander C. Berg and Li Fei-Fei},
Title = { {ImageNet Large Scale Visual Recognition Challenge} },
Year = {2015},
journal = {International Journal of Computer Vision (IJCV)},
doi = {10.1007/s11263-015-0816-y},
volume={115},
number={3},
pages={211-252}
}