Multi-Head Self-Attention via Vision Transformer for Zero-Shot Learning

Relying on massive annotated datasets, significant progress has been made on many visual recognition tasks, which is mainly due to the widespread use