- תיאור :
מערך הנתונים של ה-Quick Draw הוא אוסף של 50 מיליון ציורים על פני 345 קטגוריות, שנתרמו על ידי שחקני המשחק Quick, Draw!. מערך הנתונים של מפת סיביות מכיל ציורים אלה שהומרו מפורמט וקטור לתמונות בגווני אפור בגודל 28x28
תיעוד נוסף : חקור על ניירות עם קוד
דף הבית : https://github.com/googlecreativelab/quickdraw-dataset
קוד מקור :
tfds.datasets.quickdraw_bitmap.Builderגרסאות :
-
3.0.0(ברירת מחדל): API חדש מפוצל ( https://tensorflow.org/datasets/splits )
-
גודל הורדה :
36.82 GiBגודל ערכת נתונים :
Unknown sizeשמור אוטומטי במטמון ( תיעוד ): לא ידוע
פיצולים :
| לְפַצֵל | דוגמאות |
|---|---|
'train' | 50,426,266 |
- מבנה תכונה :
FeaturesDict({
'image': Image(shape=(28, 28, 1), dtype=uint8),
'label': ClassLabel(shape=(), dtype=int64, num_classes=345),
})
- תיעוד תכונה :
| תכונה | מעמד | צוּרָה | Dtype | תיאור |
|---|---|---|---|---|
| FeaturesDict | ||||
| תמונה | תמונה | (28, 28, 1) | uint8 | |
| תווית | ClassLabel | int64 |
מפתחות בפיקוח (ראה
as_superviseddoc ):('image', 'label')איור ( tfds.show_examples ):

- דוגמאות ( tfds.as_dataframe ):
- ציטוט :
@article{DBLP:journals/corr/HaE17,
author = {David Ha and
Douglas Eck},
title = {A Neural Representation of Sketch Drawings},
journal = {CoRR},
volume = {abs/1704.03477},
year = {2017},
url = {http://arxiv.org/abs/1704.03477},
archivePrefix = {arXiv},
eprint = {1704.03477},
timestamp = {Mon, 13 Aug 2018 16:48:30 +0200},
biburl = {https://dblp.org/rec/bib/journals/corr/HaE17},
bibsource = {dblp computer science bibliography, https://dblp.org}
}