- Descrizione :
ImageNet-PI è una versione rietichettata del dataset standard ILSVRC2012 ImageNet, in cui le etichette sono fornite da una raccolta di 16 reti neurali profonde con diverse architetture pre-addestrate sullo standard ILSVRC2012. Nello specifico, i modelli pre-addestrati vengono scaricati da tf.keras.applications.
Oltre alle nuove etichette, ImageNet-PI fornisce anche metadati sul processo di annotazione sotto forma di confidenze dei modelli sulle loro etichette e informazioni aggiuntive su ciascun modello.
Per maggiori informazioni vedere: ImageNet-PI
Pagina iniziale : https://github.com/google-research-datasets/imagenet_pi/
Codice sorgente :
tfds.datasets.imagenet_pi.BuilderVersioni :
-
1.0.0(predefinita): versione iniziale.
-
Dimensione download :
Unknown sizeDimensione del set di dati :
Unknown sizeIstruzioni per il download manuale : questo set di dati richiede di scaricare manualmente i dati di origine in
download_config.manual_dir(il valore predefinito è~/tensorflow_datasets/downloads/manual/):
manual_dir dovrebbe contenere due file: ILSVRC2012_img_train.tar e ILSVRC2012_img_val.tar. È necessario registrarsi su http://www.image-net.org/download-images per ottenere il link per scaricare il dataset.Memorizzato automaticamente nella cache ( documentazione ): Sconosciuto
Divisioni :
| Diviso | Esempi |
|---|
- Struttura delle caratteristiche :
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),
})
- Documentazione delle funzionalità :
| Caratteristica | Classe | Forma | Tipo D | Descrizione |
|---|---|---|---|---|
| CaratteristicheDict | ||||
| annotatore_confidenze | Tensore | (16,) | galleggiante32 | |
| etichette_annotatore | Tensore | (16,) | int64 | |
| etichetta pulita | Etichetta di classe | int64 | ||
| nome_file | Testo | corda | ||
| immagine | Immagine | (Nessuno, Nessuno, 3) | uint8 |
Chiavi supervisionate (vedere
as_superviseddoc ):('image', 'annotator_labels')Figura ( tfds.show_examples ): Non supportato.
Esempi ( tfds.as_dataframe ): Mancanti.
Citazione :
@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}
}