open_images_v4

Open Images হল ~9M ইমেজের একটি ডেটাসেট যা ইমেজ-লেভেল লেবেল এবং অবজেক্ট বাউন্ডিং বাক্সের সাথে টীকা করা হয়েছে।

V4-এর প্রশিক্ষণ সেটে 1.74M ছবিতে 600টি অবজেক্ট ক্লাসের জন্য 14.6M বাউন্ডিং বক্স রয়েছে, যা এটিকে অবজেক্টের অবস্থান টীকা সহ বিদ্যমান বৃহত্তম ডেটাসেট করে তুলেছে। সঠিকতা এবং ধারাবাহিকতা নিশ্চিত করার জন্য বাক্সগুলি মূলত পেশাদার টীকাকারদের দ্বারা ম্যানুয়ালি আঁকা হয়েছে। চিত্রগুলি খুব বৈচিত্র্যময় এবং প্রায়শই বেশ কয়েকটি বস্তু সহ জটিল দৃশ্য ধারণ করে (প্রতি চিত্র গড়ে 8.4)। অধিকন্তু, ডেটাসেটটি হাজার হাজার শ্রেণীতে বিস্তৃত চিত্র-স্তরের লেবেলগুলির সাথে টীকা করা হয়েছে।

বিভক্ত উদাহরণ
'test' 125,436
'train' 1,743,042
'validation' 41,620
  • বৈশিষ্ট্য গঠন :
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),
    }),
})
  • বৈশিষ্ট্য ডকুমেন্টেশন :
বৈশিষ্ট্য ক্লাস আকৃতি ডিটাইপ বর্ণনা
ফিচারসডিক্ট
ববজেক্ট ক্রম
bobjects/bbox বিবক্স ফিচার (৪,) float32
bobjects/is_depiction টেনসর int8
bobjects/is_group_of টেনসর int8
bobjects/is_inside টেনসর int8
bobjects/is_occluded টেনসর int8
bobjects/is_truncated টেনসর int8
ববজেক্ট/লেবেল ক্লাসলেবেল int64
ববজেক্ট/উৎস ক্লাসলেবেল int64
ইমেজ ছবি (কোনটিই নয়, 3) uint8
ছবি/ফাইলের নাম পাঠ্য স্ট্রিং
বস্তু ক্রম
বস্তু/আস্থা টেনসর int32
বস্তু/লেবেল ক্লাসলেবেল int64
বস্তু/উৎস ক্লাসলেবেল int64
বস্তু_প্রশিক্ষণযোগ্য ক্রম
বস্তু_প্রশিক্ষণযোগ্য/আত্মবিশ্বাস টেনসর int32
বস্তু_প্রশিক্ষণযোগ্য/লেবেল ক্লাসলেবেল int64
বস্তু_প্রশিক্ষণযোগ্য/উৎস ক্লাসলেবেল int64
  • তত্ত্বাবধান করা কী (দেখুন as_supervised doc ): None

  • উদ্ধৃতি :

@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/অরিজিনাল (ডিফল্ট কনফিগারেশন)

  • কনফিগারেশনের বিবরণ : ছবিগুলি তাদের আসল রেজোলিউশন এবং মানের।

  • ডেটাসেটের আকার : 562.42 GiB

  • চিত্র ( tfds.show_examples ):

ভিজ্যুয়ালাইজেশন

open_images_v4/300k

  • কনফিগারেশনের বিবরণ : 72 JPEG মানের চিত্রে প্রায় 300,000 পিক্সেল আছে।

  • ডেটাসেটের আকার : 81.92 GiB

  • চিত্র ( tfds.show_examples ):

ভিজ্যুয়ালাইজেশন

open_images_v4/200k

  • কনফিগারেশনের বিবরণ : 72 JPEG মানের চিত্রে মোটামুটি 200,000 পিক্সেল আছে।

  • ডেটাসেটের আকার : 60.70 GiB

  • চিত্র ( tfds.show_examples ):

ভিজ্যুয়ালাইজেশন