Open GokulNC opened 3 years ago
As in: Contrastive Self-Supervised Learning for Skeleton Action Recognition
Viewpoint Augmentations:
RotationTransform
ScaleNormalization
ScaleTransform
As in: Contrastive Self-Supervised Learning for Skeleton Action Recognition
Viewpoint Augmentations:
RotationTransform
only does rotation about 1 axis. Extend to 2 axis (if possible, 3D)ScaleNormalization
before sending to network, is this necessary?)ScaleTransform
should handle this?)