Standard convnets do not contain explicit rotational invariance (unless you include a layer such as this: arXiv:1506.02025v1). They can however learn rotational

Standard convnets do not contain explicit rotational invariance (unless you include a layer such as this: arXiv:1506.02025v1). They can however learn rotational invariance if you feed them rotated images.