NeurIPS-2021 Intriguing Properties of Vision Transformers
Properties of ViT — skim-level notes
- Because ViT adopts a transformer backbone, it can aggregate global context early; standard convolutional networks usually need many layers before they effectively capture image-wide information.
- Improved robustness: stronger performance under occlusion, distribution shift, adversarial attacks, permuted patch order, and in semantic segmentation settings.
<HR align=left color=#987cb9 SIZE=1>