smallnorb

This database is intended for experiments in 3D object recognition from shape. It contains images of 50 toys belonging to 5 generic categories: four-legged animals, human figures, airplanes, trucks, and cars. The objects were imaged by two cameras under 6 lighting conditions, 9 elevations (30 to 70 degrees every 5 degrees), and 18 azimuths (0 to 340 every 20 degrees).

The training set is composed of 5 instances of each category (instances 4, 6, 7, 8 and 9), and the test set of the remaining 5 instances (instances 0, 1, 2, 3, and 5).

Split Examples
'test' 24,300
'train' 24,300
  • Feature structure:
FeaturesDict({
    'image': Image(shape=(96, 96, 1), dtype=uint8),
    'image2': Image(shape=(96, 96, 1), dtype=uint8),
    'instance': ClassLabel(shape=(), dtype=int64, num_classes=10),
    'label_azimuth': ClassLabel(shape=(), dtype=int64, num_classes=18),
    'label_category': ClassLabel(shape=(), dtype=int64, num_classes=5),
    'label_elevation': ClassLabel(shape=(), dtype=int64, num_classes=9),
    'label_lighting': ClassLabel(shape=(), dtype=int64, num_classes=6),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
image Image (96, 96, 1) uint8
image2 Image (96, 96, 1) uint8
instance ClassLabel int64
label_azimuth ClassLabel int64
label_category ClassLabel int64
label_elevation ClassLabel int64
label_lighting ClassLabel int64
  • Citation:
@article{LeCun2004LearningMF,
  title={Learning methods for generic object recognition with invariance to pose and lighting},
  author={Yann LeCun and Fu Jie Huang and L{\'e}on Bottou},
  journal={Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition},
  year={2004},
  volume={2},
  pages={II-104 Vol.2}
}