Vision Transformers (ViT)

Attention Map Visualization

Interpretability technique visualizing attention weights between patches to understand which image regions the model focuses on.

← Wstecz