Robustness of Deep Learning Models for Vision Tasks

In recent years, artificial intelligence technologies in vision tasks have gradually begun to be applied the physical world, proving they are vulnerable adversarial attacks. Thus, importance of improving robustness against attacks has emerged as an urgent issue tasks. This article aims provide a historical summary evolution and defense methods on CNN-based models also introduces studies focusing brain-inspired that mimic visual cortex, which is resistant As origination CNN was application physiological findings related cortex time, new opportunity create more robust The authors hope this review will promote interest progress artificially intelligent security by deep learning for