inteligente

  • Descripción :

CLEVR es un conjunto de datos de diagnóstico que prueba una variedad de habilidades de razonamiento visual. Contiene sesgos mínimos y tiene anotaciones detalladas que describen el tipo de razonamiento que requiere cada pregunta.

Dividir Ejemplos
'test' 15.000
'train' 70.000
'validation' 15.000
  • Estructura de características :
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),
   
}),
})
  • Documentación de funciones :
Característica Clase Forma tipo D Descripción
FuncionesDict
Nombre del archivo Texto cadena
imagen Imagen (Ninguno, Ninguno, 3) uint8
objetos Secuencia
objetos/3d_coords Tensor (3,) flotador32
objetos/color Etiqueta de clase int64
objetos/materiales Etiqueta de clase int64
objetos/coords_píxeles Tensor (3,) flotador32
objetos/rotación Tensor flotador32
objetos/forma Etiqueta de clase int64
objetos/tamaño Etiqueta de clase int64
pregunta respuesta Secuencia
pregunta_respuesta/respuesta Texto cadena
pregunta_respuesta/pregunta Texto cadena

Visualización

  • Citación :
@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}
}