coco_captions

  • Description :

COCO est un ensemble de données de détection, de segmentation et de sous-titrage d'objets à grande échelle. Cette version contient des images, des cadres de délimitation, des étiquettes et des légendes de COCO 2014, divisés en sous-ensembles définis par Karpathy et Li (2015). Cela divise efficacement les données de validation originales du COCO 2014 en nouveaux ensembles de validation et de test de 5 000 images, plus un ensemble « restval » contenant les ~ 30 000 images restantes. Toutes les divisions ont des annotations de légende.

Diviser Exemples
'restval' 30 504
'test' 5 000
'train' 82 783
'val' 5 000
  • Structure des fonctionnalités :
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),
    }),
})
  • Documentation des fonctionnalités :
Fonctionnalité Classe Forme Type D Description
FonctionnalitésDict
légendes Séquence
légendes/identifiant Tenseur int64
légendes/texte Tenseur chaîne
image Image (Aucun, Aucun, 3) uint8
image/nom de fichier Texte chaîne
image/identifiant Tenseur int64
objets Séquence
objets/zone Tenseur int64
objets/bbox Fonctionnalité BBox (4,) flotteur32
objets/identifiant Tenseur int64
objets/is_crowd Tenseur bouffon
objets/étiquette Étiquette de classe int64

Visualisation

  • Citation :
@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 (configuration par défaut)