What is attention mechanism in vision models?
Updated May 15, 2026
Short answer
Attention allows models to focus on important regions of an image.
Deep explanation
Attention computes weighted importance of different spatial regions or feature tokens. It enhances relevant features while suppressing irrelevant ones. Self-attention compares all positions in a feature map.
Real-world example
Used in vision transformers for image classification.
Common mistakes
- Assuming attention is only for NLP tasks.
Follow-up questions
- What is self-attention?
- Why is attention powerful?