intelligente

CLEVR è un set di dati diagnostici che mette alla prova una serie di capacità di ragionamento visivo. Contiene pregiudizi minimi e ha annotazioni dettagliate che descrivono il tipo di ragionamento richiesto da ciascuna domanda.

Diviso Esempi
'test' 15.000
'train' 70.000
'validation' 15.000
  • Struttura delle caratteristiche :
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),
    }),
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
nome del file Testo corda
Immagine Immagine (Nessuno, Nessuno, 3) uint8
oggetti Sequenza
oggetti/coordinate_3d Tensore (3,) galleggiante32
oggetti/colore ClassLabel int64
oggetti/materiale ClassLabel int64
oggetti/pixel_coords Tensore (3,) galleggiante32
oggetti/rotazione Tensore galleggiante32
oggetti/forma ClassLabel int64
oggetti/dimensioni ClassLabel int64
domanda risposta Sequenza
domanda_risposta/risposta Testo corda
domanda_risposta/domanda Testo corda

Visualizzazione

  • Citazione :
@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}
}