אוקספורד_iiit_pet

  • תיאור :

מערך הנתונים של אוקספורד-IIIT לחיות מחמד הוא מערך תמונות של חיות מחמד בקטגוריות של 37 עם בערך 200 תמונות לכל מחלקה. לתמונות יש וריאציות גדולות בקנה מידה, תנוחה ותאורה. לכל התמונות יש הערת אמת קרקע קשורה של גזע ומינים. בנוסף, מסופקות תיבות תוחמות ראש עבור פיצול האימון, המאפשרות שימוש במערך נתונים זה למשימות זיהוי אובייקטים פשוטות. בפיצול הבדיקה, התיבות התוחמות ריקות.

לְפַצֵל דוגמאות
'test' 3,669
'train' 3,680
  • מבנה תכונה :
FeaturesDict({
    'file_name': Text(shape=(), dtype=string),
    'head_bbox': BBoxFeature(shape=(4,), dtype=float32),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=37),
    'segmentation_mask': Image(shape=(None, None, 1), dtype=uint8),
    'species': ClassLabel(shape=(), dtype=int64, num_classes=2),
})
  • תיעוד תכונה :
תכונה מַחלָקָה צוּרָה Dtype תֵאוּר
FeaturesDict
file_name טֶקסט חוּט
head_bbox BBoxFeature (4,) לצוף32
תְמוּנָה תְמוּנָה (אין, אין, 3) uint8
מַדבֵּקָה ClassLabel int64
segmentation_mask תְמוּנָה (אין, אין, 1) uint8
מִין ClassLabel int64
@InProceedings{parkhi12a,
  author       = "Parkhi, O. M. and Vedaldi, A. and Zisserman, A. and Jawahar, C.~V.",
  title        = "Cats and Dogs",
  booktitle    = "IEEE Conference on Computer Vision and Pattern Recognition",
  year         = "2012",
}