coco_captions

COCO는 대규모 개체 감지, 세분화 및 캡션 데이터 세트입니다. 이 버전에는 COCO 2014의 이미지, 경계 상자, 레이블 및 캡션이 포함되어 있으며 Karpathy 및 Li(2015)에서 정의한 하위 집합으로 나뉩니다. 이렇게 하면 원본 COCO 2014 유효성 검사 데이터를 새로운 5000개 이미지 유효성 검사 및 테스트 세트와 나머지 ~30k 이미지가 포함된 "restval" 세트로 효과적으로 나눕니다. 모든 분할에는 캡션 주석이 있습니다.

나뉘다
'restval' 30,504
'test' 5,000
'train' 82,783
'val' 5,000
  • 기능 구조 :
FeaturesDict({
    'captions': Sequence({
        'id': int64,
        'text': string,
    }),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'image/filename': Text(shape=(), dtype=string),
    'image/id': int64,
    'objects': Sequence({
        'area': int64,
        'bbox': BBoxFeature(shape=(4,), dtype=float32),
        'id': int64,
        'is_crowd': bool,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=80),
    }),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
캡션 순서
캡션/ID 텐서 int64
캡션/텍스트 텐서
영상 영상 (없음, 없음, 3) uint8
이미지/파일 이름 텍스트
이미지/ID 텐서 int64
사물 순서
개체/영역 텐서 int64
개체/bbox BBox기능 (4,) float32
객체/ID 텐서 int64
객체/is_crowd 텐서 부울
개체/레이블 클래스 레이블 int64

심상

  • 인용 :
@article{DBLP:journals/corr/LinMBHPRDZ14,
  author    = {Tsung{-}Yi Lin and
               Michael Maire and
               Serge J. Belongie and
               Lubomir D. Bourdev and
               Ross B. Girshick and
               James Hays and
               Pietro Perona and
               Deva Ramanan and
               Piotr Doll{'{a} }r and
               C. Lawrence Zitnick},
  title     = {Microsoft {COCO:} Common Objects in Context},
  journal   = {CoRR},
  volume    = {abs/1405.0312},
  year      = {2014},
  url       = {http://arxiv.org/abs/1405.0312},
  archivePrefix = {arXiv},
  eprint    = {1405.0312},
  timestamp = {Mon, 13 Aug 2018 16:48:13 +0200},
  biburl    = {https://dblp.org/rec/bib/journals/corr/LinMBHPRDZ14},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}@inproceedings{DBLP:conf/cvpr/KarpathyL15,
  author    = {Andrej Karpathy and
               Fei{-}Fei Li},
  title     = {Deep visual-semantic alignments for generating image
               descriptions},
  booktitle = { {IEEE} Conference on Computer Vision and Pattern Recognition,
               {CVPR} 2015, Boston, MA, USA, June 7-12, 2015},
  pages     = {3128--3137},
  publisher = { {IEEE} Computer Society},
  year      = {2015},
  url       = {https://doi.org/10.1109/CVPR.2015.7298932},
  doi       = {10.1109/CVPR.2015.7298932},
  timestamp = {Wed, 16 Oct 2019 14:14:50 +0200},
  biburl    = {https://dblp.org/rec/conf/cvpr/KarpathyL15.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

coco_captions/2014(기본 구성)