open_images_v4

Open Images مجموعه‌ای از 9 میلیون تصویر است که با برچسب‌های سطح تصویر و جعبه‌های محدودکننده اشیا حاشیه‌نویسی شده‌اند.

مجموعه آموزشی V4 شامل 14.6M جعبه محدود کننده برای 600 کلاس شی در 1.74M تصاویر است که آن را به بزرگترین مجموعه داده موجود با حاشیه نویسی مکان شی تبدیل می کند. جعبه ها عمدتاً به صورت دستی توسط حاشیه نویسان حرفه ای ترسیم شده اند تا از دقت و سازگاری اطمینان حاصل شود. تصاویر بسیار متنوع هستند و اغلب شامل صحنه های پیچیده با چندین شی هستند (به طور متوسط ​​8.4 در هر تصویر). علاوه بر این، مجموعه داده با برچسب‌های سطح تصویر که هزاران کلاس را در بر می‌گیرد، حاشیه‌نویسی می‌شود.

شکاف مثال ها
'test' 125,436
'train' 1,743,042
'validation' 41620
  • ساختار ویژگی :
FeaturesDict({
    'bobjects': Sequence({
        'bbox': BBoxFeature(shape=(4,), dtype=float32),
        'is_depiction': int8,
        'is_group_of': int8,
        'is_inside': int8,
        'is_occluded': int8,
        'is_truncated': int8,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=601),
        'source': ClassLabel(shape=(), dtype=int64, num_classes=6),
    }),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'image/filename': Text(shape=(), dtype=string),
    'objects': Sequence({
        'confidence': int32,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=19995),
        'source': ClassLabel(shape=(), dtype=int64, num_classes=6),
    }),
    'objects_trainable': Sequence({
        'confidence': int32,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=7186),
        'source': ClassLabel(shape=(), dtype=int64, num_classes=6),
    }),
})
  • مستندات ویژگی :
ویژگی کلاس شکل نوع D شرح
FeaturesDict
بابجکت ها توالی
bobjects/bbox ویژگی BBox (4،) float32
bobjects/is_depiction تانسور int8
bobjects/is_group_of تانسور int8
bobjects/is_inside تانسور int8
bobjects/is_closed تانسور int8
bobjects/is_truncated تانسور int8
bobjects/برچسب ClassLabel int64
bobjects/منبع ClassLabel int64
تصویر تصویر (هیچ، هیچ، 3) uint8
تصویر/نام فایل متن رشته
اشیاء توالی
اشیاء / اعتماد به نفس تانسور int32
اشیاء/برچسب ClassLabel int64
اشیاء/منبع ClassLabel int64
اشیاء_آموزش پذیر توالی
اشیاء_آموزش پذیر/اطمینان تانسور int32
objects_trainable/label ClassLabel int64
objects_trainable/source ClassLabel int64
  • کلیدهای نظارت شده (به as_supervised doc مراجعه کنید): None

  • نقل قول :

@article{OpenImages,
  author = {Alina Kuznetsova and
            Hassan Rom and
            Neil Alldrin and
            Jasper Uijlings and
            Ivan Krasin and
            Jordi Pont-Tuset and
            Shahab Kamali and
            Stefan Popov and
            Matteo Malloci and
            Tom Duerig and
            Vittorio Ferrari},
  title = {The Open Images Dataset V4: Unified image classification,
           object detection, and visual relationship detection at scale},
  year = {2018},
  journal = {arXiv:1811.00982}
}
@article{OpenImages2,
  author = {Krasin, Ivan and
            Duerig, Tom and
            Alldrin, Neil and
            Ferrari, Vittorio
            and Abu-El-Haija, Sami and
            Kuznetsova, Alina and
            Rom, Hassan and
            Uijlings, Jasper and
            Popov, Stefan and
            Kamali, Shahab and
            Malloci, Matteo and
            Pont-Tuset, Jordi and
            Veit, Andreas and
            Belongie, Serge and
            Gomes, Victor and
            Gupta, Abhinav and
            Sun, Chen and
            Chechik, Gal and
            Cai, David and
            Feng, Zheyun and
            Narayanan, Dhyanesh and
            Murphy, Kevin},
  title = {OpenImages: A public dataset for large-scale multi-label and
           multi-class image classification.},
  journal = {Dataset available from
             https://storage.googleapis.com/openimages/web/index.html},
  year={2017}
}

open_images_v4/original (پیکربندی پیش‌فرض)

  • توضیحات پیکربندی : تصاویر با وضوح و کیفیت اصلی خود.

  • حجم مجموعه داده : 562.42 GiB

  • شکل ( tfds.show_examples ):

تجسم

open_images_v4/300k

  • توضیحات پیکربندی : تصاویر تقریباً 300000 پیکسل با کیفیت 72 JPEG دارند.

  • حجم مجموعه داده : 81.92 GiB

  • شکل ( tfds.show_examples ):

تجسم

open_images_v4/200k

  • توضیحات پیکربندی : تصاویر تقریباً 200000 پیکسل با کیفیت 72 JPEG دارند.

  • حجم مجموعه داده : 60.70 GiB

  • شکل ( tfds.show_examples ):

تجسم