open_images_v4

  • 설명 :

Open Images는 이미지 수준 레이블과 객체 경계 상자로 주석이 달린 ~9M 이미지의 데이터 세트입니다.

V4의 트레이닝 세트에는 174만 개의 이미지에 600개의 개체 클래스에 대한 1460만 개의 경계 상자가 포함되어 있어 개체 위치 주석이 포함된 기존 데이터세트 중 가장 큰 규모입니다. 상자는 정확성과 일관성을 보장하기 위해 대부분 전문 주석자가 수동으로 작성했습니다. 이미지는 매우 다양하며 여러 개체가 포함된 복잡한 장면을 포함하는 경우가 많습니다(평균 이미지당 8.4개). 또한 데이터세트에는 수천 개의 클래스에 걸쳐 있는 이미지 수준 레이블로 주석이 추가됩니다.

나뉘다
'test' 125,436
'train' 1,743,042
'validation' 41,620
  • 기능 구조 :
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),
    }),
})
  • 기능 문서 :
특징 수업 모양 Dtype 설명
특징Dict
b객체 순서
b객체/bbox B박스특징 (4,) float32
bobjects/is_description 텐서 정수8
b객체/is_group_of 텐서 정수8
bobjects/is_inside 텐서 정수8
bobjects/is_occluded 텐서 정수8
bobjects/is_truncated 텐서 정수8
객체/레이블 클래스 라벨 정수64
객체/소스 클래스 라벨 정수64
영상 영상 (없음, 없음, 3) uint8
이미지/파일 이름 텍스트
사물 순서
객체/신뢰 텐서 정수32
객체/라벨 클래스 라벨 정수64
객체/소스 클래스 라벨 정수64
객체_훈련 가능 순서
객체_훈련 가능/신뢰 텐서 정수32
객체_훈련 가능/레이블 클래스 라벨 정수64
객체_훈련 가능/소스 클래스 라벨 정수64
@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 (기본 구성)

  • 구성 설명 : 이미지를 원래 해상도와 품질로 유지합니다.

  • 데이터세트 크기 : 562.42 GiB

  • 그림 ( tfds.show_examples ):

심상

open_images_v4/300k

  • 구성 설명 : 이미지는 72 JPEG 품질에서 약 300,000픽셀을 갖습니다.

  • 데이터세트 크기 : 81.92 GiB

  • 그림 ( tfds.show_examples ):

심상

open_images_v4/200k

  • 구성 설명 : 이미지는 72 JPEG 품질에서 대략 200,000픽셀을 갖습니다.

  • 데이터세트 크기 : 60.70 GiB

  • 그림 ( tfds.show_examples ):

심상