clivr

CLEVR é um conjunto de dados de diagnóstico que testa uma variedade de habilidades de raciocínio visual. Ele contém vieses mínimos e possui anotações detalhadas que descrevem o tipo de raciocínio que cada questão requer.

Dividir Exemplos
'test' 15.000
'train' 70.000
'validation' 15.000
  • Estrutura de recursos :
FeaturesDict({
    'file_name': Text(shape=(), dtype=string),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'objects': Sequence({
        '3d_coords': Tensor(shape=(3,), dtype=float32),
        'color': ClassLabel(shape=(), dtype=int64, num_classes=8),
        'material': ClassLabel(shape=(), dtype=int64, num_classes=2),
        'pixel_coords': Tensor(shape=(3,), dtype=float32),
        'rotation': float32,
        'shape': ClassLabel(shape=(), dtype=int64, num_classes=3),
        'size': ClassLabel(shape=(), dtype=int64, num_classes=2),
    }),
    'question_answer': Sequence({
        'answer': Text(shape=(), dtype=string),
        'question': Text(shape=(), dtype=string),
    }),
})
  • Documentação do recurso:
Característica Classe Forma Tipo D Descrição
RecursosDict
nome do arquivo Texto corda
imagem Imagem (Nenhuma, Nenhuma, 3) uint8
objetos Seqüência
objetos/3d_coords tensor (3,) float32
objetos/cor ClassLabel int64
objetos/materiais ClassLabel int64
objetos/pixel_coords tensor (3,) float32
objetos/rotação tensor float32
objetos/forma ClassLabel int64
objetos/tamanho ClassLabel int64
Resposta da questão Seqüência
pergunta_resposta/resposta Texto corda
pergunta_resposta/pergunta Texto corda

Visualização

  • Citação :
@inproceedings{johnson2017clevr,
  title={ {CLEVR}: A diagnostic dataset for compositional language and elementary visual reasoning},
  author={Johnson, Justin and Hariharan, Bharath and van der Maaten, Laurens and Fei-Fei, Li and Lawrence Zitnick, C and Girshick, Ross},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2017}
}