Robustifying Vision Transformer without Retraining from Scratch by Test-Time Class-Conditional Feature Alignment.

Takeshi Kojima, Yutaka Matsuo, Yusuke Iwasawa

Browse the full IJCAI paper archive.