- विवरण :
ImageNet-PI, मानक ILSVRC2012 ImageNet डेटासेट का पुनः लेबल किया गया संस्करण है, जिसमें लेबल 16 अलग-अलग आर्किटेक्चर वाले डीप न्यूरल नेटवर्क के संग्रह द्वारा प्रदान किए जाते हैं, जिन्हें मानक ILSVRC2012 पर पहले से प्रशिक्षित किया गया है। विशेष रूप से, पूर्व-प्रशिक्षित मॉडल tf.keras.applications से डाउनलोड किए जाते हैं।
नए लेबलों के अलावा, ImageNet-PI लेबलों पर मॉडलों के विश्वास और प्रत्येक मॉडल के बारे में अतिरिक्त जानकारी के रूप में एनोटेशन प्रक्रिया के बारे में मेटा-डेटा भी प्रदान करता है।
अधिक जानकारी के लिए देखें: ImageNet-PI
होमपेज : https://github.com/google-research-datasets/imagenet_pi/
स्रोत कोड :
tfds.datasets.imagenet_pi.Builderसंस्करणों :
-
1.0.0(डिफ़ॉल्ट): प्रारंभिक रिलीज़।
-
डाउनलोड आकार :
Unknown sizeडेटासेट का आकार :
Unknown sizeमैन्युअल डाउनलोड निर्देश : इस डेटासेट के लिए आपको स्रोत डेटा को मैन्युअल रूप से
download_config.manual_dirमें डाउनलोड करना होगा (डिफ़ॉल्ट रूप से~/tensorflow_datasets/downloads/manual/पर स्थित है):
manual_dir में दो फाइलें होनी चाहिए: ILSVRC2012_img_train.tar और ILSVRC2012_img_val.tar। डेटासेट डाउनलोड करने का लिंक प्राप्त करने के लिए आपको http://www.image-net.org/download-images पर पंजीकरण करना होगा।ऑटो-कैश्ड ( दस्तावेज़ीकरण ): अज्ञात
विभाजन :
| विभाजित करना | उदाहरण |
|---|
- विशेषता संरचना :
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),
})
- फ़ीचर दस्तावेज़ीकरण :
| विशेषता | कक्षा | आकार | डीटाइप | विवरण |
|---|---|---|---|---|
| विशेषता शब्दकोश | ||||
| एनोटेटर_विश्वास | टेन्सर | (16,) | फ्लोट32 | |
| एनोटेटर_लेबल | टेन्सर | (16,) | इंट64 | |
| क्लीन_लेबल | क्लासलेबल | इंट64 | ||
| फ़ाइल_नाम | मूलपाठ | डोरी | ||
| छवि | छवि | (कोई नहीं, कोई नहीं, 3) | uint8 |
पर्यवेक्षित कुंजियाँ (
as_supervisedदस्तावेज़ देखें):('image', 'annotator_labels')चित्र ( tfds.show_examples ): समर्थित नहीं है।
उदाहरण ( tfds.as_dataframe ): अनुपलब्ध।
उद्धरण :
@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}
}