Teacher-generated spatial-attention labels boost robustness and accuracy of contrastive models.

Yushi Yao, Chang Ye, Junfeng He, Gamaleldin F. Elsayed

Browse the full CVPR paper archive.