lvis

  • وصف :

LVIS: مجموعة بيانات لتجزئة مثيلات المفردات الكبيرة.

ينقسم أمثلة
'minival' 4,809
'test' 19822
'train' 100,170
'validation' 19,809
  • هيكل الميزة :
FeaturesDict({
   
'image': Image(shape=(None, None, 3), dtype=uint8),
   
'image/id': int64,
   
'neg_category_ids': Sequence(ClassLabel(shape=(), dtype=int64, num_classes=1203)),
   
'not_exhaustive_category_ids': Sequence(ClassLabel(shape=(), dtype=int64, num_classes=1203)),
   
'objects': Sequence({
       
'area': int64,
       
'bbox': BBoxFeature(shape=(4,), dtype=float32),
       
'id': int64,
       
'label': ClassLabel(shape=(), dtype=int64, num_classes=1203),
       
'segmentation': Image(shape=(None, None, 1), dtype=uint8),
   
}),
})
  • وثائق الميزة :
ميزة فصل شكل نوع D وصف
المميزاتDict
صورة صورة (لا شيء، لا شيء، 3) uint8
الصورة/المعرف الموتر int64
Neg_category_ids التسلسل (ClassLabel) (لا أحد،) int64
not_exhaustive_category_ids التسلسل (ClassLabel) (لا أحد،) int64
أشياء تسلسل
الكائنات / المنطقة الموتر int64
الكائنات/bbox ميزة BBox (4،) float32
الكائنات/المعرف الموتر int64
الكائنات/التسمية ClassLabel int64
الكائنات/التجزئة صورة (لا شيء، لا شيء، 1) uint8

التصور

  • الاقتباس :
@inproceedings{gupta2019lvis,
  title
={ {LVIS}: A Dataset for Large Vocabulary Instance Segmentation},
  author
={Gupta, Agrim and Dollar, Piotr and Girshick, Ross},
  booktitle
={Proceedings of the {IEEE} Conference on Computer Vision and Pattern Recognition},
  year
={2019}
}