clevr

CLEVR is a diagnostic dataset that tests a range of visual reasoning abilities. It contains minimal biases and has detailed annotations describing the kind of reasoning each question requires.

• URL: https://cs.stanford.edu/people/jcjohns/clevr/
• DatasetBuilder: tfds.image.clevr.CLEVR
• Version: v1.0.0

• Versions:

  • 1.0.0 (default):
  • 3.0.0: New split API (https://tensorflow.org/datasets/splits)

• Size: 17.72 GiB

Features

FeaturesDict({
    'file_name': Text(shape=(), dtype=tf.string),
    'image': Image(shape=(None, None, 3), dtype=tf.uint8),
    'objects': Sequence({
        '3d_coords': Tensor(shape=(3,), dtype=tf.float32),
        'color': ClassLabel(shape=(), dtype=tf.int64, num_classes=8),
        'material': ClassLabel(shape=(), dtype=tf.int64, num_classes=2),
        'pixel_coords': Tensor(shape=(3,), dtype=tf.float32),
        'rotation': Tensor(shape=(), dtype=tf.float32),
        'shape': ClassLabel(shape=(), dtype=tf.int64, num_classes=3),
        'size': ClassLabel(shape=(), dtype=tf.int64, num_classes=2),
    }),
})

Statistics

Homepage

• https://cs.stanford.edu/people/jcjohns/clevr/

Citation

@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}
}